From: bellard Date: Sun, 13 Mar 2005 16:55:58 +0000 (+0000) Subject: use the generic soft float code X-Git-Tag: TizenStudio_2.0_p2.3~14883 X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=20495218834824723ef81306c6e1fd27fc3ae560;p=sdk%2Femulator%2Fqemu.git use the generic soft float code git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@1333 c046a42c-6fe2-441c-8c8c-71466251a162 --- diff --git a/target-arm/nwfpe/ARM-gcc.h b/target-arm/nwfpe/ARM-gcc.h deleted file mode 100644 index e659847..0000000 --- a/target-arm/nwfpe/ARM-gcc.h +++ /dev/null @@ -1,120 +0,0 @@ -/* -------------------------------------------------------------------------------- -The macro `BITS64' can be defined to indicate that 64-bit integer types are -supported by the compiler. -------------------------------------------------------------------------------- -*/ -#define BITS64 - -/* -------------------------------------------------------------------------------- -Each of the following `typedef's defines the most convenient type that holds -integers of at least as many bits as specified. For example, `uint8' should -be the most convenient type that can hold unsigned integers of as many as -8 bits. The `flag' type must be able to hold either a 0 or 1. For most -implementations of C, `flag', `uint8', and `int8' should all be `typedef'ed -to the same as `int'. -------------------------------------------------------------------------------- -*/ -typedef char flag; -typedef unsigned char uint8; -typedef signed char int8; -typedef int uint16; -typedef int int16; -typedef unsigned int uint32; -typedef signed int int32; -#ifdef BITS64 -typedef unsigned long long int bits64; -typedef signed long long int sbits64; -#endif - -/* -------------------------------------------------------------------------------- -Each of the following `typedef's defines a type that holds integers -of _exactly_ the number of bits specified. For instance, for most -implementation of C, `bits16' and `sbits16' should be `typedef'ed to -`unsigned short int' and `signed short int' (or `short int'), respectively. -------------------------------------------------------------------------------- -*/ -typedef unsigned char bits8; -typedef signed char sbits8; -typedef unsigned short int bits16; -typedef signed short int sbits16; -typedef unsigned int bits32; -typedef signed int sbits32; -#ifdef BITS64 -typedef unsigned long long int uint64; -typedef signed long long int int64; -#endif - -#ifdef BITS64 -/* -------------------------------------------------------------------------------- -The `LIT64' macro takes as its argument a textual integer literal and if -necessary ``marks'' the literal as having a 64-bit integer type. For -example, the Gnu C Compiler (`gcc') requires that 64-bit literals be -appended with the letters `LL' standing for `long long', which is `gcc's -name for the 64-bit integer type. Some compilers may allow `LIT64' to be -defined as the identity macro: `#define LIT64( a ) a'. -------------------------------------------------------------------------------- -*/ -#define LIT64( a ) a##LL -#endif - -/* -------------------------------------------------------------------------------- -The macro `INLINE' can be used before functions that should be inlined. If -a compiler does not support explicit inlining, this macro should be defined -to be `static'. -------------------------------------------------------------------------------- -*/ -#define INLINE extern __inline__ - - -/* For use as a GCC soft-float library we need some special function names. */ - -#ifdef __LIBFLOAT__ - -/* Some 32-bit ops can be mapped straight across by just changing the name. */ -#define float32_add __addsf3 -#define float32_sub __subsf3 -#define float32_mul __mulsf3 -#define float32_div __divsf3 -#define int32_to_float32 __floatsisf -#define float32_to_int32_round_to_zero __fixsfsi -#define float32_to_uint32_round_to_zero __fixunssfsi - -/* These ones go through the glue code. To avoid namespace pollution - we rename the internal functions too. */ -#define float32_eq ___float32_eq -#define float32_le ___float32_le -#define float32_lt ___float32_lt - -/* All the 64-bit ops have to go through the glue, so we pull the same - trick. */ -#define float64_add ___float64_add -#define float64_sub ___float64_sub -#define float64_mul ___float64_mul -#define float64_div ___float64_div -#define int32_to_float64 ___int32_to_float64 -#define float64_to_int32_round_to_zero ___float64_to_int32_round_to_zero -#define float64_to_uint32_round_to_zero ___float64_to_uint32_round_to_zero -#define float64_to_float32 ___float64_to_float32 -#define float32_to_float64 ___float32_to_float64 -#define float64_eq ___float64_eq -#define float64_le ___float64_le -#define float64_lt ___float64_lt - -#if 0 -#define float64_add __adddf3 -#define float64_sub __subdf3 -#define float64_mul __muldf3 -#define float64_div __divdf3 -#define int32_to_float64 __floatsidf -#define float64_to_int32_round_to_zero __fixdfsi -#define float64_to_uint32_round_to_zero __fixunsdfsi -#define float64_to_float32 __truncdfsf2 -#define float32_to_float64 __extendsfdf2 -#endif - -#endif diff --git a/target-arm/nwfpe/double_cpdo.c b/target-arm/nwfpe/double_cpdo.c index c44d3fc..944083a 100644 --- a/target-arm/nwfpe/double_cpdo.c +++ b/target-arm/nwfpe/double_cpdo.c @@ -53,7 +53,7 @@ unsigned int DoubleCPDO(const unsigned int opcode) switch (fpa11->fType[Fm]) { case typeSingle: - rFm = float32_to_float64(fpa11->fpreg[Fm].fSingle); + rFm = float32_to_float64(fpa11->fpreg[Fm].fSingle, &fpa11->fp_status); break; case typeDouble: @@ -79,7 +79,7 @@ unsigned int DoubleCPDO(const unsigned int opcode) switch (fpa11->fType[Fn]) { case typeSingle: - rFn = float32_to_float64(fpa11->fpreg[Fn].fSingle); + rFn = float32_to_float64(fpa11->fpreg[Fn].fSingle, &fpa11->fp_status); break; case typeDouble: @@ -96,30 +96,30 @@ unsigned int DoubleCPDO(const unsigned int opcode) { /* dyadic opcodes */ case ADF_CODE: - fpa11->fpreg[Fd].fDouble = float64_add(rFn,rFm); + fpa11->fpreg[Fd].fDouble = float64_add(rFn,rFm, &fpa11->fp_status); break; case MUF_CODE: case FML_CODE: - fpa11->fpreg[Fd].fDouble = float64_mul(rFn,rFm); + fpa11->fpreg[Fd].fDouble = float64_mul(rFn,rFm, &fpa11->fp_status); break; case SUF_CODE: - fpa11->fpreg[Fd].fDouble = float64_sub(rFn,rFm); + fpa11->fpreg[Fd].fDouble = float64_sub(rFn,rFm, &fpa11->fp_status); break; case RSF_CODE: - fpa11->fpreg[Fd].fDouble = float64_sub(rFm,rFn); + fpa11->fpreg[Fd].fDouble = float64_sub(rFm,rFn, &fpa11->fp_status); break; case DVF_CODE: case FDV_CODE: - fpa11->fpreg[Fd].fDouble = float64_div(rFn,rFm); + fpa11->fpreg[Fd].fDouble = float64_div(rFn,rFm, &fpa11->fp_status); break; case RDF_CODE: case FRD_CODE: - fpa11->fpreg[Fd].fDouble = float64_div(rFm,rFn); + fpa11->fpreg[Fd].fDouble = float64_div(rFm,rFn, &fpa11->fp_status); break; #if 0 @@ -133,7 +133,7 @@ unsigned int DoubleCPDO(const unsigned int opcode) #endif case RMF_CODE: - fpa11->fpreg[Fd].fDouble = float64_rem(rFn,rFm); + fpa11->fpreg[Fd].fDouble = float64_rem(rFn,rFm, &fpa11->fp_status); break; #if 0 @@ -173,11 +173,11 @@ unsigned int DoubleCPDO(const unsigned int opcode) case RND_CODE: case URD_CODE: - fpa11->fpreg[Fd].fDouble = float64_round_to_int(rFm); + fpa11->fpreg[Fd].fDouble = float64_round_to_int(rFm, &fpa11->fp_status); break; case SQT_CODE: - fpa11->fpreg[Fd].fDouble = float64_sqrt(rFm); + fpa11->fpreg[Fd].fDouble = float64_sqrt(rFm, &fpa11->fp_status); break; #if 0 diff --git a/target-arm/nwfpe/extended_cpdo.c b/target-arm/nwfpe/extended_cpdo.c index 3314075..f5ef623 100644 --- a/target-arm/nwfpe/extended_cpdo.c +++ b/target-arm/nwfpe/extended_cpdo.c @@ -53,11 +53,11 @@ unsigned int ExtendedCPDO(const unsigned int opcode) switch (fpa11->fType[Fm]) { case typeSingle: - rFm = float32_to_floatx80(fpa11->fpreg[Fm].fSingle); + rFm = float32_to_floatx80(fpa11->fpreg[Fm].fSingle, &fpa11->fp_status); break; case typeDouble: - rFm = float64_to_floatx80(fpa11->fpreg[Fm].fDouble); + rFm = float64_to_floatx80(fpa11->fpreg[Fm].fDouble, &fpa11->fp_status); break; case typeExtended: @@ -74,11 +74,11 @@ unsigned int ExtendedCPDO(const unsigned int opcode) switch (fpa11->fType[Fn]) { case typeSingle: - rFn = float32_to_floatx80(fpa11->fpreg[Fn].fSingle); + rFn = float32_to_floatx80(fpa11->fpreg[Fn].fSingle, &fpa11->fp_status); break; case typeDouble: - rFn = float64_to_floatx80(fpa11->fpreg[Fn].fDouble); + rFn = float64_to_floatx80(fpa11->fpreg[Fn].fDouble, &fpa11->fp_status); break; case typeExtended: @@ -94,30 +94,30 @@ unsigned int ExtendedCPDO(const unsigned int opcode) { /* dyadic opcodes */ case ADF_CODE: - fpa11->fpreg[Fd].fExtended = floatx80_add(rFn,rFm); + fpa11->fpreg[Fd].fExtended = floatx80_add(rFn,rFm, &fpa11->fp_status); break; case MUF_CODE: case FML_CODE: - fpa11->fpreg[Fd].fExtended = floatx80_mul(rFn,rFm); + fpa11->fpreg[Fd].fExtended = floatx80_mul(rFn,rFm, &fpa11->fp_status); break; case SUF_CODE: - fpa11->fpreg[Fd].fExtended = floatx80_sub(rFn,rFm); + fpa11->fpreg[Fd].fExtended = floatx80_sub(rFn,rFm, &fpa11->fp_status); break; case RSF_CODE: - fpa11->fpreg[Fd].fExtended = floatx80_sub(rFm,rFn); + fpa11->fpreg[Fd].fExtended = floatx80_sub(rFm,rFn, &fpa11->fp_status); break; case DVF_CODE: case FDV_CODE: - fpa11->fpreg[Fd].fExtended = floatx80_div(rFn,rFm); + fpa11->fpreg[Fd].fExtended = floatx80_div(rFn,rFm, &fpa11->fp_status); break; case RDF_CODE: case FRD_CODE: - fpa11->fpreg[Fd].fExtended = floatx80_div(rFm,rFn); + fpa11->fpreg[Fd].fExtended = floatx80_div(rFm,rFn, &fpa11->fp_status); break; #if 0 @@ -131,7 +131,7 @@ unsigned int ExtendedCPDO(const unsigned int opcode) #endif case RMF_CODE: - fpa11->fpreg[Fd].fExtended = floatx80_rem(rFn,rFm); + fpa11->fpreg[Fd].fExtended = floatx80_rem(rFn,rFm, &fpa11->fp_status); break; #if 0 @@ -157,11 +157,11 @@ unsigned int ExtendedCPDO(const unsigned int opcode) case RND_CODE: case URD_CODE: - fpa11->fpreg[Fd].fExtended = floatx80_round_to_int(rFm); + fpa11->fpreg[Fd].fExtended = floatx80_round_to_int(rFm, &fpa11->fp_status); break; case SQT_CODE: - fpa11->fpreg[Fd].fExtended = floatx80_sqrt(rFm); + fpa11->fpreg[Fd].fExtended = floatx80_sqrt(rFm, &fpa11->fp_status); break; #if 0 diff --git a/target-arm/nwfpe/fpa11.c b/target-arm/nwfpe/fpa11.c index 143bcd3..cfbe700 100644 --- a/target-arm/nwfpe/fpa11.c +++ b/target-arm/nwfpe/fpa11.c @@ -61,74 +61,79 @@ void resetFPA11(void) void SetRoundingMode(const unsigned int opcode) { -#if MAINTAIN_FPCR + int rounding_mode; FPA11 *fpa11 = GET_FPA11(); + +#if MAINTAIN_FPCR fpa11->fpcr &= ~MASK_ROUNDING_MODE; #endif switch (opcode & MASK_ROUNDING_MODE) { default: case ROUND_TO_NEAREST: - float_rounding_mode = float_round_nearest_even; + rounding_mode = float_round_nearest_even; #if MAINTAIN_FPCR fpa11->fpcr |= ROUND_TO_NEAREST; #endif break; case ROUND_TO_PLUS_INFINITY: - float_rounding_mode = float_round_up; + rounding_mode = float_round_up; #if MAINTAIN_FPCR fpa11->fpcr |= ROUND_TO_PLUS_INFINITY; #endif break; case ROUND_TO_MINUS_INFINITY: - float_rounding_mode = float_round_down; + rounding_mode = float_round_down; #if MAINTAIN_FPCR fpa11->fpcr |= ROUND_TO_MINUS_INFINITY; #endif break; case ROUND_TO_ZERO: - float_rounding_mode = float_round_to_zero; + rounding_mode = float_round_to_zero; #if MAINTAIN_FPCR fpa11->fpcr |= ROUND_TO_ZERO; #endif break; } + set_float_rounding_mode(rounding_mode, &fpa11->fp_status); } void SetRoundingPrecision(const unsigned int opcode) { -#if MAINTAIN_FPCR + int rounding_precision; FPA11 *fpa11 = GET_FPA11(); +#if MAINTAIN_FPCR fpa11->fpcr &= ~MASK_ROUNDING_PRECISION; #endif switch (opcode & MASK_ROUNDING_PRECISION) { case ROUND_SINGLE: - floatx80_rounding_precision = 32; + rounding_precision = 32; #if MAINTAIN_FPCR fpa11->fpcr |= ROUND_SINGLE; #endif break; case ROUND_DOUBLE: - floatx80_rounding_precision = 64; + rounding_precision = 64; #if MAINTAIN_FPCR fpa11->fpcr |= ROUND_DOUBLE; #endif break; case ROUND_EXTENDED: - floatx80_rounding_precision = 80; + rounding_precision = 80; #if MAINTAIN_FPCR fpa11->fpcr |= ROUND_EXTENDED; #endif break; - default: floatx80_rounding_precision = 80; + default: rounding_precision = 80; } + set_floatx80_rounding_precision(rounding_precision, &fpa11->fp_status); } /* Emulate the instruction in the opcode. */ diff --git a/target-arm/nwfpe/fpa11.h b/target-arm/nwfpe/fpa11.h index 389c029..668393c 100644 --- a/target-arm/nwfpe/fpa11.h +++ b/target-arm/nwfpe/fpa11.h @@ -83,6 +83,7 @@ typedef struct tagFPA11 { so we can use it to detect whether this instance of the emulator needs to be initialised. */ + float_status fp_status; /* QEMU float emulator status */ } FPA11; extern FPA11* qemufpa; diff --git a/target-arm/nwfpe/fpa11_cpdo.c b/target-arm/nwfpe/fpa11_cpdo.c index 343a6b9..cc8aa87 100644 --- a/target-arm/nwfpe/fpa11_cpdo.c +++ b/target-arm/nwfpe/fpa11_cpdo.c @@ -80,10 +80,10 @@ unsigned int EmulateCPDO(const unsigned int opcode) { if (typeDouble == nType) fpa11->fpreg[Fd].fSingle = - float64_to_float32(fpa11->fpreg[Fd].fDouble); + float64_to_float32(fpa11->fpreg[Fd].fDouble, &fpa11->fp_status); else fpa11->fpreg[Fd].fSingle = - floatx80_to_float32(fpa11->fpreg[Fd].fExtended); + floatx80_to_float32(fpa11->fpreg[Fd].fExtended, &fpa11->fp_status); } break; @@ -91,10 +91,10 @@ unsigned int EmulateCPDO(const unsigned int opcode) { if (typeSingle == nType) fpa11->fpreg[Fd].fDouble = - float32_to_float64(fpa11->fpreg[Fd].fSingle); + float32_to_float64(fpa11->fpreg[Fd].fSingle, &fpa11->fp_status); else fpa11->fpreg[Fd].fDouble = - floatx80_to_float64(fpa11->fpreg[Fd].fExtended); + floatx80_to_float64(fpa11->fpreg[Fd].fExtended, &fpa11->fp_status); } break; @@ -102,10 +102,10 @@ unsigned int EmulateCPDO(const unsigned int opcode) { if (typeSingle == nType) fpa11->fpreg[Fd].fExtended = - float32_to_floatx80(fpa11->fpreg[Fd].fSingle); + float32_to_floatx80(fpa11->fpreg[Fd].fSingle, &fpa11->fp_status); else fpa11->fpreg[Fd].fExtended = - float64_to_floatx80(fpa11->fpreg[Fd].fDouble); + float64_to_floatx80(fpa11->fpreg[Fd].fDouble, &fpa11->fp_status); } break; } diff --git a/target-arm/nwfpe/fpa11_cpdt.c b/target-arm/nwfpe/fpa11_cpdt.c index c1d5615..3319d88 100644 --- a/target-arm/nwfpe/fpa11_cpdt.c +++ b/target-arm/nwfpe/fpa11_cpdt.c @@ -106,11 +106,11 @@ void storeSingle(const unsigned int Fn,unsigned int *pMem) switch (fpa11->fType[Fn]) { case typeDouble: - val = float64_to_float32(fpa11->fpreg[Fn].fDouble); + val = float64_to_float32(fpa11->fpreg[Fn].fDouble, &fpa11->fp_status); break; case typeExtended: - val = floatx80_to_float32(fpa11->fpreg[Fn].fExtended); + val = floatx80_to_float32(fpa11->fpreg[Fn].fExtended, &fpa11->fp_status); break; default: val = fpa11->fpreg[Fn].fSingle; @@ -129,11 +129,11 @@ void storeDouble(const unsigned int Fn,unsigned int *pMem) switch (fpa11->fType[Fn]) { case typeSingle: - val = float32_to_float64(fpa11->fpreg[Fn].fSingle); + val = float32_to_float64(fpa11->fpreg[Fn].fSingle, &fpa11->fp_status); break; case typeExtended: - val = floatx80_to_float64(fpa11->fpreg[Fn].fExtended); + val = floatx80_to_float64(fpa11->fpreg[Fn].fExtended, &fpa11->fp_status); break; default: val = fpa11->fpreg[Fn].fDouble; @@ -157,11 +157,11 @@ void storeExtended(const unsigned int Fn,unsigned int *pMem) switch (fpa11->fType[Fn]) { case typeSingle: - val = float32_to_floatx80(fpa11->fpreg[Fn].fSingle); + val = float32_to_floatx80(fpa11->fpreg[Fn].fSingle, &fpa11->fp_status); break; case typeDouble: - val = float64_to_floatx80(fpa11->fpreg[Fn].fDouble); + val = float64_to_floatx80(fpa11->fpreg[Fn].fDouble, &fpa11->fp_status); break; default: val = fpa11->fpreg[Fn].fExtended; diff --git a/target-arm/nwfpe/fpa11_cprt.c b/target-arm/nwfpe/fpa11_cprt.c index 17871c1..fe295e1 100644 --- a/target-arm/nwfpe/fpa11_cprt.c +++ b/target-arm/nwfpe/fpa11_cprt.c @@ -21,7 +21,6 @@ */ #include "fpa11.h" -#include "milieu.h" #include "softfloat.h" #include "fpopcode.h" #include "fpa11.inl" @@ -89,7 +88,7 @@ unsigned int PerformFLT(const unsigned int opcode) { fpa11->fType[getFn(opcode)] = typeSingle; fpa11->fpreg[getFn(opcode)].fSingle = - int32_to_float32(readRegister(getRd(opcode))); + int32_to_float32(readRegister(getRd(opcode)), &fpa11->fp_status); } break; @@ -97,7 +96,7 @@ unsigned int PerformFLT(const unsigned int opcode) { fpa11->fType[getFn(opcode)] = typeDouble; fpa11->fpreg[getFn(opcode)].fDouble = - int32_to_float64(readRegister(getRd(opcode))); + int32_to_float64(readRegister(getRd(opcode)), &fpa11->fp_status); } break; @@ -105,7 +104,7 @@ unsigned int PerformFLT(const unsigned int opcode) { fpa11->fType[getFn(opcode)] = typeExtended; fpa11->fpreg[getFn(opcode)].fExtended = - int32_to_floatx80(readRegister(getRd(opcode))); + int32_to_floatx80(readRegister(getRd(opcode)), &fpa11->fp_status); } break; @@ -128,7 +127,7 @@ unsigned int PerformFIX(const unsigned int opcode) case typeSingle: { writeRegister(getRd(opcode), - float32_to_int32(fpa11->fpreg[Fn].fSingle)); + float32_to_int32(fpa11->fpreg[Fn].fSingle, &fpa11->fp_status)); } break; @@ -136,14 +135,14 @@ unsigned int PerformFIX(const unsigned int opcode) { //printf("F%d is 0x%llx\n",Fn,fpa11->fpreg[Fn].fDouble); writeRegister(getRd(opcode), - float64_to_int32(fpa11->fpreg[Fn].fDouble)); + float64_to_int32(fpa11->fpreg[Fn].fDouble, &fpa11->fp_status)); } break; case typeExtended: { writeRegister(getRd(opcode), - floatx80_to_int32(fpa11->fpreg[Fn].fExtended)); + floatx80_to_int32(fpa11->fpreg[Fn].fExtended, &fpa11->fp_status)); } break; @@ -157,22 +156,23 @@ unsigned int PerformFIX(const unsigned int opcode) static unsigned int __inline__ PerformComparisonOperation(floatx80 Fn, floatx80 Fm) { + FPA11 *fpa11 = GET_FPA11(); unsigned int flags = 0; /* test for less than condition */ - if (floatx80_lt(Fn,Fm)) + if (floatx80_lt(Fn,Fm, &fpa11->fp_status)) { flags |= CC_NEGATIVE; } /* test for equal condition */ - if (floatx80_eq(Fn,Fm)) + if (floatx80_eq(Fn,Fm, &fpa11->fp_status)) { flags |= CC_ZERO; } /* test for greater than or equal condition */ - if (floatx80_lt(Fm,Fn)) + if (floatx80_lt(Fm,Fn, &fpa11->fp_status)) { flags |= CC_CARRY; } @@ -208,14 +208,14 @@ static unsigned int PerformComparison(const unsigned int opcode) //printk("single.\n"); if (float32_is_nan(fpa11->fpreg[Fn].fSingle)) goto unordered; - rFn = float32_to_floatx80(fpa11->fpreg[Fn].fSingle); + rFn = float32_to_floatx80(fpa11->fpreg[Fn].fSingle, &fpa11->fp_status); break; case typeDouble: //printk("double.\n"); if (float64_is_nan(fpa11->fpreg[Fn].fDouble)) goto unordered; - rFn = float64_to_floatx80(fpa11->fpreg[Fn].fDouble); + rFn = float64_to_floatx80(fpa11->fpreg[Fn].fDouble, &fpa11->fp_status); break; case typeExtended: @@ -244,14 +244,14 @@ static unsigned int PerformComparison(const unsigned int opcode) //printk("single.\n"); if (float32_is_nan(fpa11->fpreg[Fm].fSingle)) goto unordered; - rFm = float32_to_floatx80(fpa11->fpreg[Fm].fSingle); + rFm = float32_to_floatx80(fpa11->fpreg[Fm].fSingle, &fpa11->fp_status); break; case typeDouble: //printk("double.\n"); if (float64_is_nan(fpa11->fpreg[Fm].fDouble)) goto unordered; - rFm = float64_to_floatx80(fpa11->fpreg[Fm].fDouble); + rFm = float64_to_floatx80(fpa11->fpreg[Fm].fDouble, &fpa11->fp_status); break; case typeExtended: @@ -283,7 +283,7 @@ static unsigned int PerformComparison(const unsigned int opcode) if (BIT_AC & readFPSR()) flags |= CC_CARRY; - if (e_flag) float_raise(float_flag_invalid); + if (e_flag) float_raise(float_flag_invalid, &fpa11->fp_status); writeConditionCodes(flags); return 1; diff --git a/target-arm/nwfpe/fpopcode.c b/target-arm/nwfpe/fpopcode.c index 0886a0b..d29e913 100644 --- a/target-arm/nwfpe/fpopcode.c +++ b/target-arm/nwfpe/fpopcode.c @@ -27,14 +27,14 @@ //#include "fpmodule.inl" const floatx80 floatx80Constant[] = { - { 0x0000, 0x0000000000000000ULL}, /* extended 0.0 */ - { 0x3fff, 0x8000000000000000ULL}, /* extended 1.0 */ - { 0x4000, 0x8000000000000000ULL}, /* extended 2.0 */ - { 0x4000, 0xc000000000000000ULL}, /* extended 3.0 */ - { 0x4001, 0x8000000000000000ULL}, /* extended 4.0 */ - { 0x4001, 0xa000000000000000ULL}, /* extended 5.0 */ - { 0x3ffe, 0x8000000000000000ULL}, /* extended 0.5 */ - { 0x4002, 0xa000000000000000ULL} /* extended 10.0 */ + { 0x0000000000000000ULL, 0x0000}, /* extended 0.0 */ + { 0x8000000000000000ULL, 0x3fff}, /* extended 1.0 */ + { 0x8000000000000000ULL, 0x4000}, /* extended 2.0 */ + { 0xc000000000000000ULL, 0x4000}, /* extended 3.0 */ + { 0x8000000000000000ULL, 0x4001}, /* extended 4.0 */ + { 0xa000000000000000ULL, 0x4001}, /* extended 5.0 */ + { 0x8000000000000000ULL, 0x3ffe}, /* extended 0.5 */ + { 0xa000000000000000ULL, 0x4002} /* extended 10.0 */ }; const float64 float64Constant[] = { diff --git a/target-arm/nwfpe/milieu.h b/target-arm/nwfpe/milieu.h deleted file mode 100644 index a3892ab..0000000 --- a/target-arm/nwfpe/milieu.h +++ /dev/null @@ -1,48 +0,0 @@ - -/* -=============================================================================== - -This C header file is part of the SoftFloat IEC/IEEE Floating-point -Arithmetic Package, Release 2. - -Written by John R. Hauser. This work was made possible in part by the -International Computer Science Institute, located at Suite 600, 1947 Center -Street, Berkeley, California 94704. Funding was partially provided by the -National Science Foundation under grant MIP-9311980. The original version -of this code was written as part of a project to build a fixed-point vector -processor in collaboration with the University of California at Berkeley, -overseen by Profs. Nelson Morgan and John Wawrzynek. More information -is available through the Web page `http://HTTP.CS.Berkeley.EDU/~jhauser/ -arithmetic/softfloat.html'. - -THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort -has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT -TIMES RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO -PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY -AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE. - -Derivative works are acceptable, even for commercial purposes, so long as -(1) they include prominent notice that the work is derivative, and (2) they -include prominent notice akin to these three paragraphs for those parts of -this code that are retained. - -=============================================================================== -*/ - -/* -------------------------------------------------------------------------------- -Include common integer types and flags. -------------------------------------------------------------------------------- -*/ -#include "ARM-gcc.h" - -/* -------------------------------------------------------------------------------- -Symbolic Boolean literals. -------------------------------------------------------------------------------- -*/ -enum { - FALSE = 0, - TRUE = 1 -}; - diff --git a/target-arm/nwfpe/single_cpdo.c b/target-arm/nwfpe/single_cpdo.c index 58da89b..7dd2620 100644 --- a/target-arm/nwfpe/single_cpdo.c +++ b/target-arm/nwfpe/single_cpdo.c @@ -76,30 +76,30 @@ unsigned int SingleCPDO(const unsigned int opcode) { /* dyadic opcodes */ case ADF_CODE: - fpa11->fpreg[Fd].fSingle = float32_add(rFn,rFm); + fpa11->fpreg[Fd].fSingle = float32_add(rFn,rFm, &fpa11->fp_status); break; case MUF_CODE: case FML_CODE: - fpa11->fpreg[Fd].fSingle = float32_mul(rFn,rFm); + fpa11->fpreg[Fd].fSingle = float32_mul(rFn,rFm, &fpa11->fp_status); break; case SUF_CODE: - fpa11->fpreg[Fd].fSingle = float32_sub(rFn,rFm); + fpa11->fpreg[Fd].fSingle = float32_sub(rFn,rFm, &fpa11->fp_status); break; case RSF_CODE: - fpa11->fpreg[Fd].fSingle = float32_sub(rFm,rFn); + fpa11->fpreg[Fd].fSingle = float32_sub(rFm,rFn, &fpa11->fp_status); break; case DVF_CODE: case FDV_CODE: - fpa11->fpreg[Fd].fSingle = float32_div(rFn,rFm); + fpa11->fpreg[Fd].fSingle = float32_div(rFn,rFm, &fpa11->fp_status); break; case RDF_CODE: case FRD_CODE: - fpa11->fpreg[Fd].fSingle = float32_div(rFm,rFn); + fpa11->fpreg[Fd].fSingle = float32_div(rFm,rFn, &fpa11->fp_status); break; #if 0 @@ -113,7 +113,7 @@ unsigned int SingleCPDO(const unsigned int opcode) #endif case RMF_CODE: - fpa11->fpreg[Fd].fSingle = float32_rem(rFn,rFm); + fpa11->fpreg[Fd].fSingle = float32_rem(rFn,rFm, &fpa11->fp_status); break; #if 0 @@ -139,11 +139,11 @@ unsigned int SingleCPDO(const unsigned int opcode) case RND_CODE: case URD_CODE: - fpa11->fpreg[Fd].fSingle = float32_round_to_int(rFm); + fpa11->fpreg[Fd].fSingle = float32_round_to_int(rFm, &fpa11->fp_status); break; case SQT_CODE: - fpa11->fpreg[Fd].fSingle = float32_sqrt(rFm); + fpa11->fpreg[Fd].fSingle = float32_sqrt(rFm, &fpa11->fp_status); break; #if 0 diff --git a/target-arm/nwfpe/softfloat-macros b/target-arm/nwfpe/softfloat-macros deleted file mode 100644 index c245a0e..0000000 --- a/target-arm/nwfpe/softfloat-macros +++ /dev/null @@ -1,740 +0,0 @@ - -/* -=============================================================================== - -This C source fragment is part of the SoftFloat IEC/IEEE Floating-point -Arithmetic Package, Release 2. - -Written by John R. Hauser. This work was made possible in part by the -International Computer Science Institute, located at Suite 600, 1947 Center -Street, Berkeley, California 94704. Funding was partially provided by the -National Science Foundation under grant MIP-9311980. The original version -of this code was written as part of a project to build a fixed-point vector -processor in collaboration with the University of California at Berkeley, -overseen by Profs. Nelson Morgan and John Wawrzynek. More information -is available through the web page `http://HTTP.CS.Berkeley.EDU/~jhauser/ -arithmetic/softfloat.html'. - -THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort -has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT -TIMES RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO -PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY -AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE. - -Derivative works are acceptable, even for commercial purposes, so long as -(1) they include prominent notice that the work is derivative, and (2) they -include prominent notice akin to these three paragraphs for those parts of -this code that are retained. - -=============================================================================== -*/ - -/* -------------------------------------------------------------------------------- -Shifts `a' right by the number of bits given in `count'. If any nonzero -bits are shifted off, they are ``jammed'' into the least significant bit of -the result by setting the least significant bit to 1. The value of `count' -can be arbitrarily large; in particular, if `count' is greater than 32, the -result will be either 0 or 1, depending on whether `a' is zero or nonzero. -The result is stored in the location pointed to by `zPtr'. -------------------------------------------------------------------------------- -*/ -INLINE void shift32RightJamming( bits32 a, int16 count, bits32 *zPtr ) -{ - bits32 z; - if ( count == 0 ) { - z = a; - } - else if ( count < 32 ) { - z = ( a>>count ) | ( ( a<<( ( - count ) & 31 ) ) != 0 ); - } - else { - z = ( a != 0 ); - } - *zPtr = z; -} - -/* -------------------------------------------------------------------------------- -Shifts `a' right by the number of bits given in `count'. If any nonzero -bits are shifted off, they are ``jammed'' into the least significant bit of -the result by setting the least significant bit to 1. The value of `count' -can be arbitrarily large; in particular, if `count' is greater than 64, the -result will be either 0 or 1, depending on whether `a' is zero or nonzero. -The result is stored in the location pointed to by `zPtr'. -------------------------------------------------------------------------------- -*/ -INLINE void shift64RightJamming( bits64 a, int16 count, bits64 *zPtr ) -{ - bits64 z; - -// __asm__("@shift64RightJamming -- start"); - if ( count == 0 ) { - z = a; - } - else if ( count < 64 ) { - z = ( a>>count ) | ( ( a<<( ( - count ) & 63 ) ) != 0 ); - } - else { - z = ( a != 0 ); - } -// __asm__("@shift64RightJamming -- end"); - *zPtr = z; -} - -/* -------------------------------------------------------------------------------- -Shifts the 128-bit value formed by concatenating `a0' and `a1' right by 64 -_plus_ the number of bits given in `count'. The shifted result is at most -64 nonzero bits; this is stored at the location pointed to by `z0Ptr'. The -bits shifted off form a second 64-bit result as follows: The _last_ bit -shifted off is the most-significant bit of the extra result, and the other -63 bits of the extra result are all zero if and only if _all_but_the_last_ -bits shifted off were all zero. This extra result is stored in the location -pointed to by `z1Ptr'. The value of `count' can be arbitrarily large. - (This routine makes more sense if `a0' and `a1' are considered to form a -fixed-point value with binary point between `a0' and `a1'. This fixed-point -value is shifted right by the number of bits given in `count', and the -integer part of the result is returned at the location pointed to by -`z0Ptr'. The fractional part of the result may be slightly corrupted as -described above, and is returned at the location pointed to by `z1Ptr'.) -------------------------------------------------------------------------------- -*/ -INLINE void - shift64ExtraRightJamming( - bits64 a0, bits64 a1, int16 count, bits64 *z0Ptr, bits64 *z1Ptr ) -{ - bits64 z0, z1; - int8 negCount = ( - count ) & 63; - - if ( count == 0 ) { - z1 = a1; - z0 = a0; - } - else if ( count < 64 ) { - z1 = ( a0<>count; - } - else { - if ( count == 64 ) { - z1 = a0 | ( a1 != 0 ); - } - else { - z1 = ( ( a0 | a1 ) != 0 ); - } - z0 = 0; - } - *z1Ptr = z1; - *z0Ptr = z0; - -} - -/* -------------------------------------------------------------------------------- -Shifts the 128-bit value formed by concatenating `a0' and `a1' right by the -number of bits given in `count'. Any bits shifted off are lost. The value -of `count' can be arbitrarily large; in particular, if `count' is greater -than 128, the result will be 0. The result is broken into two 64-bit pieces -which are stored at the locations pointed to by `z0Ptr' and `z1Ptr'. -------------------------------------------------------------------------------- -*/ -INLINE void - shift128Right( - bits64 a0, bits64 a1, int16 count, bits64 *z0Ptr, bits64 *z1Ptr ) -{ - bits64 z0, z1; - int8 negCount = ( - count ) & 63; - - if ( count == 0 ) { - z1 = a1; - z0 = a0; - } - else if ( count < 64 ) { - z1 = ( a0<>count ); - z0 = a0>>count; - } - else { - z1 = ( count < 64 ) ? ( a0>>( count & 63 ) ) : 0; - z0 = 0; - } - *z1Ptr = z1; - *z0Ptr = z0; - -} - -/* -------------------------------------------------------------------------------- -Shifts the 128-bit value formed by concatenating `a0' and `a1' right by the -number of bits given in `count'. If any nonzero bits are shifted off, they -are ``jammed'' into the least significant bit of the result by setting the -least significant bit to 1. The value of `count' can be arbitrarily large; -in particular, if `count' is greater than 128, the result will be either 0 -or 1, depending on whether the concatenation of `a0' and `a1' is zero or -nonzero. The result is broken into two 64-bit pieces which are stored at -the locations pointed to by `z0Ptr' and `z1Ptr'. -------------------------------------------------------------------------------- -*/ -INLINE void - shift128RightJamming( - bits64 a0, bits64 a1, int16 count, bits64 *z0Ptr, bits64 *z1Ptr ) -{ - bits64 z0, z1; - int8 negCount = ( - count ) & 63; - - if ( count == 0 ) { - z1 = a1; - z0 = a0; - } - else if ( count < 64 ) { - z1 = ( a0<>count ) | ( ( a1<>count; - } - else { - if ( count == 64 ) { - z1 = a0 | ( a1 != 0 ); - } - else if ( count < 128 ) { - z1 = ( a0>>( count & 63 ) ) | ( ( ( a0<>count ); - z0 = a0>>count; - } - else { - if ( count == 64 ) { - z2 = a1; - z1 = a0; - } - else { - a2 |= a1; - if ( count < 128 ) { - z2 = a0<>( count & 63 ); - } - else { - z2 = ( count == 128 ) ? a0 : ( a0 != 0 ); - z1 = 0; - } - } - z0 = 0; - } - z2 |= ( a2 != 0 ); - } - *z2Ptr = z2; - *z1Ptr = z1; - *z0Ptr = z0; - -} - -/* -------------------------------------------------------------------------------- -Shifts the 128-bit value formed by concatenating `a0' and `a1' left by the -number of bits given in `count'. Any bits shifted off are lost. The value -of `count' must be less than 64. The result is broken into two 64-bit -pieces which are stored at the locations pointed to by `z0Ptr' and `z1Ptr'. -------------------------------------------------------------------------------- -*/ -INLINE void - shortShift128Left( - bits64 a0, bits64 a1, int16 count, bits64 *z0Ptr, bits64 *z1Ptr ) -{ - - *z1Ptr = a1<>( ( - count ) & 63 ) ); - -} - -/* -------------------------------------------------------------------------------- -Shifts the 192-bit value formed by concatenating `a0', `a1', and `a2' left -by the number of bits given in `count'. Any bits shifted off are lost. -The value of `count' must be less than 64. The result is broken into three -64-bit pieces which are stored at the locations pointed to by `z0Ptr', -`z1Ptr', and `z2Ptr'. -------------------------------------------------------------------------------- -*/ -INLINE void - shortShift192Left( - bits64 a0, - bits64 a1, - bits64 a2, - int16 count, - bits64 *z0Ptr, - bits64 *z1Ptr, - bits64 *z2Ptr - ) -{ - bits64 z0, z1, z2; - int8 negCount; - - z2 = a2<>negCount; - z0 |= a1>>negCount; - } - *z2Ptr = z2; - *z1Ptr = z1; - *z0Ptr = z0; - -} - -/* -------------------------------------------------------------------------------- -Adds the 128-bit value formed by concatenating `a0' and `a1' to the 128-bit -value formed by concatenating `b0' and `b1'. Addition is modulo 2^128, so -any carry out is lost. The result is broken into two 64-bit pieces which -are stored at the locations pointed to by `z0Ptr' and `z1Ptr'. -------------------------------------------------------------------------------- -*/ -INLINE void - add128( - bits64 a0, bits64 a1, bits64 b0, bits64 b1, bits64 *z0Ptr, bits64 *z1Ptr ) -{ - bits64 z1; - - z1 = a1 + b1; - *z1Ptr = z1; - *z0Ptr = a0 + b0 + ( z1 < a1 ); - -} - -/* -------------------------------------------------------------------------------- -Adds the 192-bit value formed by concatenating `a0', `a1', and `a2' to the -192-bit value formed by concatenating `b0', `b1', and `b2'. Addition is -modulo 2^192, so any carry out is lost. The result is broken into three -64-bit pieces which are stored at the locations pointed to by `z0Ptr', -`z1Ptr', and `z2Ptr'. -------------------------------------------------------------------------------- -*/ -INLINE void - add192( - bits64 a0, - bits64 a1, - bits64 a2, - bits64 b0, - bits64 b1, - bits64 b2, - bits64 *z0Ptr, - bits64 *z1Ptr, - bits64 *z2Ptr - ) -{ - bits64 z0, z1, z2; - int8 carry0, carry1; - - z2 = a2 + b2; - carry1 = ( z2 < a2 ); - z1 = a1 + b1; - carry0 = ( z1 < a1 ); - z0 = a0 + b0; - z1 += carry1; - z0 += ( z1 < carry1 ); - z0 += carry0; - *z2Ptr = z2; - *z1Ptr = z1; - *z0Ptr = z0; - -} - -/* -------------------------------------------------------------------------------- -Subtracts the 128-bit value formed by concatenating `b0' and `b1' from the -128-bit value formed by concatenating `a0' and `a1'. Subtraction is modulo -2^128, so any borrow out (carry out) is lost. The result is broken into two -64-bit pieces which are stored at the locations pointed to by `z0Ptr' and -`z1Ptr'. -------------------------------------------------------------------------------- -*/ -INLINE void - sub128( - bits64 a0, bits64 a1, bits64 b0, bits64 b1, bits64 *z0Ptr, bits64 *z1Ptr ) -{ - - *z1Ptr = a1 - b1; - *z0Ptr = a0 - b0 - ( a1 < b1 ); - -} - -/* -------------------------------------------------------------------------------- -Subtracts the 192-bit value formed by concatenating `b0', `b1', and `b2' -from the 192-bit value formed by concatenating `a0', `a1', and `a2'. -Subtraction is modulo 2^192, so any borrow out (carry out) is lost. The -result is broken into three 64-bit pieces which are stored at the locations -pointed to by `z0Ptr', `z1Ptr', and `z2Ptr'. -------------------------------------------------------------------------------- -*/ -INLINE void - sub192( - bits64 a0, - bits64 a1, - bits64 a2, - bits64 b0, - bits64 b1, - bits64 b2, - bits64 *z0Ptr, - bits64 *z1Ptr, - bits64 *z2Ptr - ) -{ - bits64 z0, z1, z2; - int8 borrow0, borrow1; - - z2 = a2 - b2; - borrow1 = ( a2 < b2 ); - z1 = a1 - b1; - borrow0 = ( a1 < b1 ); - z0 = a0 - b0; - z0 -= ( z1 < borrow1 ); - z1 -= borrow1; - z0 -= borrow0; - *z2Ptr = z2; - *z1Ptr = z1; - *z0Ptr = z0; - -} - -/* -------------------------------------------------------------------------------- -Multiplies `a' by `b' to obtain a 128-bit product. The product is broken -into two 64-bit pieces which are stored at the locations pointed to by -`z0Ptr' and `z1Ptr'. -------------------------------------------------------------------------------- -*/ -INLINE void mul64To128( bits64 a, bits64 b, bits64 *z0Ptr, bits64 *z1Ptr ) -{ - bits32 aHigh, aLow, bHigh, bLow; - bits64 z0, zMiddleA, zMiddleB, z1; - - aLow = a; - aHigh = a>>32; - bLow = b; - bHigh = b>>32; - z1 = ( (bits64) aLow ) * bLow; - zMiddleA = ( (bits64) aLow ) * bHigh; - zMiddleB = ( (bits64) aHigh ) * bLow; - z0 = ( (bits64) aHigh ) * bHigh; - zMiddleA += zMiddleB; - z0 += ( ( (bits64) ( zMiddleA < zMiddleB ) )<<32 ) + ( zMiddleA>>32 ); - zMiddleA <<= 32; - z1 += zMiddleA; - z0 += ( z1 < zMiddleA ); - *z1Ptr = z1; - *z0Ptr = z0; - -} - -/* -------------------------------------------------------------------------------- -Multiplies the 128-bit value formed by concatenating `a0' and `a1' by `b' to -obtain a 192-bit product. The product is broken into three 64-bit pieces -which are stored at the locations pointed to by `z0Ptr', `z1Ptr', and -`z2Ptr'. -------------------------------------------------------------------------------- -*/ -INLINE void - mul128By64To192( - bits64 a0, - bits64 a1, - bits64 b, - bits64 *z0Ptr, - bits64 *z1Ptr, - bits64 *z2Ptr - ) -{ - bits64 z0, z1, z2, more1; - - mul64To128( a1, b, &z1, &z2 ); - mul64To128( a0, b, &z0, &more1 ); - add128( z0, more1, 0, z1, &z0, &z1 ); - *z2Ptr = z2; - *z1Ptr = z1; - *z0Ptr = z0; - -} - -/* -------------------------------------------------------------------------------- -Multiplies the 128-bit value formed by concatenating `a0' and `a1' to the -128-bit value formed by concatenating `b0' and `b1' to obtain a 256-bit -product. The product is broken into four 64-bit pieces which are stored at -the locations pointed to by `z0Ptr', `z1Ptr', `z2Ptr', and `z3Ptr'. -------------------------------------------------------------------------------- -*/ -INLINE void - mul128To256( - bits64 a0, - bits64 a1, - bits64 b0, - bits64 b1, - bits64 *z0Ptr, - bits64 *z1Ptr, - bits64 *z2Ptr, - bits64 *z3Ptr - ) -{ - bits64 z0, z1, z2, z3; - bits64 more1, more2; - - mul64To128( a1, b1, &z2, &z3 ); - mul64To128( a1, b0, &z1, &more2 ); - add128( z1, more2, 0, z2, &z1, &z2 ); - mul64To128( a0, b0, &z0, &more1 ); - add128( z0, more1, 0, z1, &z0, &z1 ); - mul64To128( a0, b1, &more1, &more2 ); - add128( more1, more2, 0, z2, &more1, &z2 ); - add128( z0, z1, 0, more1, &z0, &z1 ); - *z3Ptr = z3; - *z2Ptr = z2; - *z1Ptr = z1; - *z0Ptr = z0; - -} - -/* -------------------------------------------------------------------------------- -Returns an approximation to the 64-bit integer quotient obtained by dividing -`b' into the 128-bit value formed by concatenating `a0' and `a1'. The -divisor `b' must be at least 2^63. If q is the exact quotient truncated -toward zero, the approximation returned lies between q and q + 2 inclusive. -If the exact quotient q is larger than 64 bits, the maximum positive 64-bit -unsigned integer is returned. -------------------------------------------------------------------------------- -*/ -static bits64 estimateDiv128To64( bits64 a0, bits64 a1, bits64 b ) -{ - bits64 b0, b1; - bits64 rem0, rem1, term0, term1; - bits64 z; - if ( b <= a0 ) return LIT64( 0xFFFFFFFFFFFFFFFF ); - b0 = b>>32; - z = ( b0<<32 <= a0 ) ? LIT64( 0xFFFFFFFF00000000 ) : ( a0 / b0 )<<32; - mul64To128( b, z, &term0, &term1 ); - sub128( a0, a1, term0, term1, &rem0, &rem1 ); - while ( ( (sbits64) rem0 ) < 0 ) { - z -= LIT64( 0x100000000 ); - b1 = b<<32; - add128( rem0, rem1, b0, b1, &rem0, &rem1 ); - } - rem0 = ( rem0<<32 ) | ( rem1>>32 ); - z |= ( b0<<32 <= rem0 ) ? 0xFFFFFFFF : rem0 / b0; - return z; - -} - -/* -------------------------------------------------------------------------------- -Returns an approximation to the square root of the 32-bit significand given -by `a'. Considered as an integer, `a' must be at least 2^31. If bit 0 of -`aExp' (the least significant bit) is 1, the integer returned approximates -2^31*sqrt(`a'/2^31), where `a' is considered an integer. If bit 0 of `aExp' -is 0, the integer returned approximates 2^31*sqrt(`a'/2^30). In either -case, the approximation returned lies strictly within +/-2 of the exact -value. -------------------------------------------------------------------------------- -*/ -static bits32 estimateSqrt32( int16 aExp, bits32 a ) -{ - static const bits16 sqrtOddAdjustments[] = { - 0x0004, 0x0022, 0x005D, 0x00B1, 0x011D, 0x019F, 0x0236, 0x02E0, - 0x039C, 0x0468, 0x0545, 0x0631, 0x072B, 0x0832, 0x0946, 0x0A67 - }; - static const bits16 sqrtEvenAdjustments[] = { - 0x0A2D, 0x08AF, 0x075A, 0x0629, 0x051A, 0x0429, 0x0356, 0x029E, - 0x0200, 0x0179, 0x0109, 0x00AF, 0x0068, 0x0034, 0x0012, 0x0002 - }; - int8 index; - bits32 z; - - index = ( a>>27 ) & 15; - if ( aExp & 1 ) { - z = 0x4000 + ( a>>17 ) - sqrtOddAdjustments[ index ]; - z = ( ( a / z )<<14 ) + ( z<<15 ); - a >>= 1; - } - else { - z = 0x8000 + ( a>>17 ) - sqrtEvenAdjustments[ index ]; - z = a / z + z; - z = ( 0x20000 <= z ) ? 0xFFFF8000 : ( z<<15 ); - if ( z <= a ) return (bits32) ( ( (sbits32) a )>>1 ); - } - return ( (bits32) ( ( ( (bits64) a )<<31 ) / z ) ) + ( z>>1 ); - -} - -/* -------------------------------------------------------------------------------- -Returns the number of leading 0 bits before the most-significant 1 bit -of `a'. If `a' is zero, 32 is returned. -------------------------------------------------------------------------------- -*/ -static int8 countLeadingZeros32( bits32 a ) -{ - static const int8 countLeadingZerosHigh[] = { - 8, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4, - 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, - 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, - 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 - }; - int8 shiftCount; - - shiftCount = 0; - if ( a < 0x10000 ) { - shiftCount += 16; - a <<= 16; - } - if ( a < 0x1000000 ) { - shiftCount += 8; - a <<= 8; - } - shiftCount += countLeadingZerosHigh[ a>>24 ]; - return shiftCount; - -} - -/* -------------------------------------------------------------------------------- -Returns the number of leading 0 bits before the most-significant 1 bit -of `a'. If `a' is zero, 64 is returned. -------------------------------------------------------------------------------- -*/ -static int8 countLeadingZeros64( bits64 a ) -{ - int8 shiftCount; - - shiftCount = 0; - if ( a < ( (bits64) 1 )<<32 ) { - shiftCount += 32; - } - else { - a >>= 32; - } - shiftCount += countLeadingZeros32( a ); - return shiftCount; - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' -is equal to the 128-bit value formed by concatenating `b0' and `b1'. -Otherwise, returns 0. -------------------------------------------------------------------------------- -*/ -INLINE flag eq128( bits64 a0, bits64 a1, bits64 b0, bits64 b1 ) -{ - - return ( a0 == b0 ) && ( a1 == b1 ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is less -than or equal to the 128-bit value formed by concatenating `b0' and `b1'. -Otherwise, returns 0. -------------------------------------------------------------------------------- -*/ -INLINE flag le128( bits64 a0, bits64 a1, bits64 b0, bits64 b1 ) -{ - - return ( a0 < b0 ) || ( ( a0 == b0 ) && ( a1 <= b1 ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is less -than the 128-bit value formed by concatenating `b0' and `b1'. Otherwise, -returns 0. -------------------------------------------------------------------------------- -*/ -INLINE flag lt128( bits64 a0, bits64 a1, bits64 b0, bits64 b1 ) -{ - - return ( a0 < b0 ) || ( ( a0 == b0 ) && ( a1 < b1 ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the 128-bit value formed by concatenating `a0' and `a1' is -not equal to the 128-bit value formed by concatenating `b0' and `b1'. -Otherwise, returns 0. -------------------------------------------------------------------------------- -*/ -INLINE flag ne128( bits64 a0, bits64 a1, bits64 b0, bits64 b1 ) -{ - - return ( a0 != b0 ) || ( a1 != b1 ); - -} - diff --git a/target-arm/nwfpe/softfloat-specialize b/target-arm/nwfpe/softfloat-specialize deleted file mode 100644 index a23a8a3..0000000 --- a/target-arm/nwfpe/softfloat-specialize +++ /dev/null @@ -1,366 +0,0 @@ - -/* -=============================================================================== - -This C source fragment is part of the SoftFloat IEC/IEEE Floating-point -Arithmetic Package, Release 2. - -Written by John R. Hauser. This work was made possible in part by the -International Computer Science Institute, located at Suite 600, 1947 Center -Street, Berkeley, California 94704. Funding was partially provided by the -National Science Foundation under grant MIP-9311980. The original version -of this code was written as part of a project to build a fixed-point vector -processor in collaboration with the University of California at Berkeley, -overseen by Profs. Nelson Morgan and John Wawrzynek. More information -is available through the Web page `http://HTTP.CS.Berkeley.EDU/~jhauser/ -arithmetic/softfloat.html'. - -THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort -has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT -TIMES RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO -PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY -AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE. - -Derivative works are acceptable, even for commercial purposes, so long as -(1) they include prominent notice that the work is derivative, and (2) they -include prominent notice akin to these three paragraphs for those parts of -this code that are retained. - -=============================================================================== -*/ - -/* -------------------------------------------------------------------------------- -Underflow tininess-detection mode, statically initialized to default value. -(The declaration in `softfloat.h' must match the `int8' type here.) -------------------------------------------------------------------------------- -*/ -int8 float_detect_tininess = float_tininess_after_rounding; - -/* -------------------------------------------------------------------------------- -Raises the exceptions specified by `flags'. Floating-point traps can be -defined here if desired. It is currently not possible for such a trap to -substitute a result value. If traps are not implemented, this routine -should be simply `float_exception_flags |= flags;'. - -ScottB: November 4, 1998 -Moved this function out of softfloat-specialize into fpmodule.c. -This effectively isolates all the changes required for integrating with the -Linux kernel into fpmodule.c. Porting to NetBSD should only require modifying -fpmodule.c to integrate with the NetBSD kernel (I hope!). -------------------------------------------------------------------------------- -*/ -void float_raise( int8 flags ) -{ - float_exception_flags |= flags; -} - -/* -------------------------------------------------------------------------------- -Internal canonical NaN format. -------------------------------------------------------------------------------- -*/ -typedef struct { - flag sign; - bits64 high, low; -} commonNaNT; - -/* -------------------------------------------------------------------------------- -The pattern for a default generated single-precision NaN. -------------------------------------------------------------------------------- -*/ -#define float32_default_nan 0xFFFFFFFF - -/* -------------------------------------------------------------------------------- -Returns 1 if the single-precision floating-point value `a' is a NaN; -otherwise returns 0. -------------------------------------------------------------------------------- -*/ -flag float32_is_nan( float32 a ) -{ - - return ( 0xFF000000 < (bits32) ( a<<1 ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the single-precision floating-point value `a' is a signaling -NaN; otherwise returns 0. -------------------------------------------------------------------------------- -*/ -flag float32_is_signaling_nan( float32 a ) -{ - - return ( ( ( a>>22 ) & 0x1FF ) == 0x1FE ) && ( a & 0x003FFFFF ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of converting the single-precision floating-point NaN -`a' to the canonical NaN format. If `a' is a signaling NaN, the invalid -exception is raised. -------------------------------------------------------------------------------- -*/ -static commonNaNT float32ToCommonNaN( float32 a ) -{ - commonNaNT z; - - if ( float32_is_signaling_nan( a ) ) float_raise( float_flag_invalid ); - z.sign = a>>31; - z.low = 0; - z.high = ( (bits64) a )<<41; - return z; - -} - -/* -------------------------------------------------------------------------------- -Returns the result of converting the canonical NaN `a' to the single- -precision floating-point format. -------------------------------------------------------------------------------- -*/ -static float32 commonNaNToFloat32( commonNaNT a ) -{ - - return ( ( (bits32) a.sign )<<31 ) | 0x7FC00000 | ( a.high>>41 ); - -} - -/* -------------------------------------------------------------------------------- -Takes two single-precision floating-point values `a' and `b', one of which -is a NaN, and returns the appropriate NaN result. If either `a' or `b' is a -signaling NaN, the invalid exception is raised. -------------------------------------------------------------------------------- -*/ -static float32 propagateFloat32NaN( float32 a, float32 b ) -{ - flag aIsNaN, aIsSignalingNaN, bIsNaN, bIsSignalingNaN; - - aIsNaN = float32_is_nan( a ); - aIsSignalingNaN = float32_is_signaling_nan( a ); - bIsNaN = float32_is_nan( b ); - bIsSignalingNaN = float32_is_signaling_nan( b ); - a |= 0x00400000; - b |= 0x00400000; - if ( aIsSignalingNaN | bIsSignalingNaN ) float_raise( float_flag_invalid ); - if ( aIsNaN ) { - return ( aIsSignalingNaN & bIsNaN ) ? b : a; - } - else { - return b; - } - -} - -/* -------------------------------------------------------------------------------- -The pattern for a default generated double-precision NaN. -------------------------------------------------------------------------------- -*/ -#define float64_default_nan LIT64( 0xFFFFFFFFFFFFFFFF ) - -/* -------------------------------------------------------------------------------- -Returns 1 if the double-precision floating-point value `a' is a NaN; -otherwise returns 0. -------------------------------------------------------------------------------- -*/ -flag float64_is_nan( float64 a ) -{ - - return ( LIT64( 0xFFE0000000000000 ) < (bits64) ( a<<1 ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the double-precision floating-point value `a' is a signaling -NaN; otherwise returns 0. -------------------------------------------------------------------------------- -*/ -flag float64_is_signaling_nan( float64 a ) -{ - - return - ( ( ( a>>51 ) & 0xFFF ) == 0xFFE ) - && ( a & LIT64( 0x0007FFFFFFFFFFFF ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of converting the double-precision floating-point NaN -`a' to the canonical NaN format. If `a' is a signaling NaN, the invalid -exception is raised. -------------------------------------------------------------------------------- -*/ -static commonNaNT float64ToCommonNaN( float64 a ) -{ - commonNaNT z; - - if ( float64_is_signaling_nan( a ) ) float_raise( float_flag_invalid ); - z.sign = a>>63; - z.low = 0; - z.high = a<<12; - return z; - -} - -/* -------------------------------------------------------------------------------- -Returns the result of converting the canonical NaN `a' to the double- -precision floating-point format. -------------------------------------------------------------------------------- -*/ -static float64 commonNaNToFloat64( commonNaNT a ) -{ - - return - ( ( (bits64) a.sign )<<63 ) - | LIT64( 0x7FF8000000000000 ) - | ( a.high>>12 ); - -} - -/* -------------------------------------------------------------------------------- -Takes two double-precision floating-point values `a' and `b', one of which -is a NaN, and returns the appropriate NaN result. If either `a' or `b' is a -signaling NaN, the invalid exception is raised. -------------------------------------------------------------------------------- -*/ -static float64 propagateFloat64NaN( float64 a, float64 b ) -{ - flag aIsNaN, aIsSignalingNaN, bIsNaN, bIsSignalingNaN; - - aIsNaN = float64_is_nan( a ); - aIsSignalingNaN = float64_is_signaling_nan( a ); - bIsNaN = float64_is_nan( b ); - bIsSignalingNaN = float64_is_signaling_nan( b ); - a |= LIT64( 0x0008000000000000 ); - b |= LIT64( 0x0008000000000000 ); - if ( aIsSignalingNaN | bIsSignalingNaN ) float_raise( float_flag_invalid ); - if ( aIsNaN ) { - return ( aIsSignalingNaN & bIsNaN ) ? b : a; - } - else { - return b; - } - -} - -#ifdef FLOATX80 - -/* -------------------------------------------------------------------------------- -The pattern for a default generated extended double-precision NaN. The -`high' and `low' values hold the most- and least-significant bits, -respectively. -------------------------------------------------------------------------------- -*/ -#define floatx80_default_nan_high 0xFFFF -#define floatx80_default_nan_low LIT64( 0xFFFFFFFFFFFFFFFF ) - -/* -------------------------------------------------------------------------------- -Returns 1 if the extended double-precision floating-point value `a' is a -NaN; otherwise returns 0. -------------------------------------------------------------------------------- -*/ -flag floatx80_is_nan( floatx80 a ) -{ - - return ( ( a.high & 0x7FFF ) == 0x7FFF ) && (bits64) ( a.low<<1 ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the extended double-precision floating-point value `a' is a -signaling NaN; otherwise returns 0. -------------------------------------------------------------------------------- -*/ -flag floatx80_is_signaling_nan( floatx80 a ) -{ - //register int lr; - bits64 aLow; - - //__asm__("mov %0, lr" : : "g" (lr)); - //fp_printk("floatx80_is_signalling_nan() called from 0x%08x\n",lr); - aLow = a.low & ~ LIT64( 0x4000000000000000 ); - return - ( ( a.high & 0x7FFF ) == 0x7FFF ) - && (bits64) ( aLow<<1 ) - && ( a.low == aLow ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of converting the extended double-precision floating- -point NaN `a' to the canonical NaN format. If `a' is a signaling NaN, the -invalid exception is raised. -------------------------------------------------------------------------------- -*/ -static commonNaNT floatx80ToCommonNaN( floatx80 a ) -{ - commonNaNT z; - - if ( floatx80_is_signaling_nan( a ) ) float_raise( float_flag_invalid ); - z.sign = a.high>>15; - z.low = 0; - z.high = a.low<<1; - return z; - -} - -/* -------------------------------------------------------------------------------- -Returns the result of converting the canonical NaN `a' to the extended -double-precision floating-point format. -------------------------------------------------------------------------------- -*/ -static floatx80 commonNaNToFloatx80( commonNaNT a ) -{ - floatx80 z; - - z.low = LIT64( 0xC000000000000000 ) | ( a.high>>1 ); - z.high = ( ( (bits16) a.sign )<<15 ) | 0x7FFF; - return z; - -} - -/* -------------------------------------------------------------------------------- -Takes two extended double-precision floating-point values `a' and `b', one -of which is a NaN, and returns the appropriate NaN result. If either `a' or -`b' is a signaling NaN, the invalid exception is raised. -------------------------------------------------------------------------------- -*/ -static floatx80 propagateFloatx80NaN( floatx80 a, floatx80 b ) -{ - flag aIsNaN, aIsSignalingNaN, bIsNaN, bIsSignalingNaN; - - aIsNaN = floatx80_is_nan( a ); - aIsSignalingNaN = floatx80_is_signaling_nan( a ); - bIsNaN = floatx80_is_nan( b ); - bIsSignalingNaN = floatx80_is_signaling_nan( b ); - a.low |= LIT64( 0xC000000000000000 ); - b.low |= LIT64( 0xC000000000000000 ); - if ( aIsSignalingNaN | bIsSignalingNaN ) float_raise( float_flag_invalid ); - if ( aIsNaN ) { - return ( aIsSignalingNaN & bIsNaN ) ? b : a; - } - else { - return b; - } - -} - -#endif diff --git a/target-arm/nwfpe/softfloat.c b/target-arm/nwfpe/softfloat.c deleted file mode 100644 index 8ffb9a9..0000000 --- a/target-arm/nwfpe/softfloat.c +++ /dev/null @@ -1,3427 +0,0 @@ -/* -=============================================================================== - -This C source file is part of the SoftFloat IEC/IEEE Floating-point -Arithmetic Package, Release 2. - -Written by John R. Hauser. This work was made possible in part by the -International Computer Science Institute, located at Suite 600, 1947 Center -Street, Berkeley, California 94704. Funding was partially provided by the -National Science Foundation under grant MIP-9311980. The original version -of this code was written as part of a project to build a fixed-point vector -processor in collaboration with the University of California at Berkeley, -overseen by Profs. Nelson Morgan and John Wawrzynek. More information -is available through the web page `http://HTTP.CS.Berkeley.EDU/~jhauser/ -arithmetic/softfloat.html'. - -THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort -has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT -TIMES RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO -PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY -AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE. - -Derivative works are acceptable, even for commercial purposes, so long as -(1) they include prominent notice that the work is derivative, and (2) they -include prominent notice akin to these three paragraphs for those parts of -this code that are retained. - -=============================================================================== -*/ - -#include "fpa11.h" -#include "milieu.h" -#include "softfloat.h" - -/* -------------------------------------------------------------------------------- -Floating-point rounding mode, extended double-precision rounding precision, -and exception flags. -------------------------------------------------------------------------------- -*/ -int8 float_rounding_mode = float_round_nearest_even; -int8 floatx80_rounding_precision = 80; -int8 float_exception_flags; - -/* -------------------------------------------------------------------------------- -Primitive arithmetic functions, including multi-word arithmetic, and -division and square root approximations. (Can be specialized to target if -desired.) -------------------------------------------------------------------------------- -*/ -#include "softfloat-macros" - -/* -------------------------------------------------------------------------------- -Functions and definitions to determine: (1) whether tininess for underflow -is detected before or after rounding by default, (2) what (if anything) -happens when exceptions are raised, (3) how signaling NaNs are distinguished -from quiet NaNs, (4) the default generated quiet NaNs, and (5) how NaNs -are propagated from function inputs to output. These details are target- -specific. -------------------------------------------------------------------------------- -*/ -#include "softfloat-specialize" - -/* -------------------------------------------------------------------------------- -Takes a 64-bit fixed-point value `absZ' with binary point between bits 6 -and 7, and returns the properly rounded 32-bit integer corresponding to the -input. If `zSign' is nonzero, the input is negated before being converted -to an integer. Bit 63 of `absZ' must be zero. Ordinarily, the fixed-point -input is simply rounded to an integer, with the inexact exception raised if -the input cannot be represented exactly as an integer. If the fixed-point -input is too large, however, the invalid exception is raised and the largest -positive or negative integer is returned. -------------------------------------------------------------------------------- -*/ -static int32 roundAndPackInt32( flag zSign, bits64 absZ ) -{ - int8 roundingMode; - flag roundNearestEven; - int8 roundIncrement, roundBits; - int32 z; - - roundingMode = float_rounding_mode; - roundNearestEven = ( roundingMode == float_round_nearest_even ); - roundIncrement = 0x40; - if ( ! roundNearestEven ) { - if ( roundingMode == float_round_to_zero ) { - roundIncrement = 0; - } - else { - roundIncrement = 0x7F; - if ( zSign ) { - if ( roundingMode == float_round_up ) roundIncrement = 0; - } - else { - if ( roundingMode == float_round_down ) roundIncrement = 0; - } - } - } - roundBits = absZ & 0x7F; - absZ = ( absZ + roundIncrement )>>7; - absZ &= ~ ( ( ( roundBits ^ 0x40 ) == 0 ) & roundNearestEven ); - z = absZ; - if ( zSign ) z = - z; - if ( ( absZ>>32 ) || ( z && ( ( z < 0 ) ^ zSign ) ) ) { - float_exception_flags |= float_flag_invalid; - return zSign ? 0x80000000 : 0x7FFFFFFF; - } - if ( roundBits ) float_exception_flags |= float_flag_inexact; - return z; - -} - -/* -------------------------------------------------------------------------------- -Returns the fraction bits of the single-precision floating-point value `a'. -------------------------------------------------------------------------------- -*/ -INLINE bits32 extractFloat32Frac( float32 a ) -{ - - return a & 0x007FFFFF; - -} - -/* -------------------------------------------------------------------------------- -Returns the exponent bits of the single-precision floating-point value `a'. -------------------------------------------------------------------------------- -*/ -INLINE int16 extractFloat32Exp( float32 a ) -{ - - return ( a>>23 ) & 0xFF; - -} - -/* -------------------------------------------------------------------------------- -Returns the sign bit of the single-precision floating-point value `a'. -------------------------------------------------------------------------------- -*/ -INLINE flag extractFloat32Sign( float32 a ) -{ - - return a>>31; - -} - -/* -------------------------------------------------------------------------------- -Normalizes the subnormal single-precision floating-point value represented -by the denormalized significand `aSig'. The normalized exponent and -significand are stored at the locations pointed to by `zExpPtr' and -`zSigPtr', respectively. -------------------------------------------------------------------------------- -*/ -static void - normalizeFloat32Subnormal( bits32 aSig, int16 *zExpPtr, bits32 *zSigPtr ) -{ - int8 shiftCount; - - shiftCount = countLeadingZeros32( aSig ) - 8; - *zSigPtr = aSig<>7; - zSig &= ~ ( ( ( roundBits ^ 0x40 ) == 0 ) & roundNearestEven ); - if ( zSig == 0 ) zExp = 0; - return packFloat32( zSign, zExp, zSig ); - -} - -/* -------------------------------------------------------------------------------- -Takes an abstract floating-point value having sign `zSign', exponent `zExp', -and significand `zSig', and returns the proper single-precision floating- -point value corresponding to the abstract input. This routine is just like -`roundAndPackFloat32' except that `zSig' does not have to be normalized in -any way. In all cases, `zExp' must be 1 less than the ``true'' floating- -point exponent. -------------------------------------------------------------------------------- -*/ -static float32 - normalizeRoundAndPackFloat32( flag zSign, int16 zExp, bits32 zSig ) -{ - int8 shiftCount; - - shiftCount = countLeadingZeros32( zSig ) - 1; - return roundAndPackFloat32( zSign, zExp - shiftCount, zSig<>52 ) & 0x7FF; - -} - -/* -------------------------------------------------------------------------------- -Returns the sign bit of the double-precision floating-point value `a'. -------------------------------------------------------------------------------- -*/ -INLINE flag extractFloat64Sign( float64 a ) -{ - - return a>>63; - -} - -/* -------------------------------------------------------------------------------- -Normalizes the subnormal double-precision floating-point value represented -by the denormalized significand `aSig'. The normalized exponent and -significand are stored at the locations pointed to by `zExpPtr' and -`zSigPtr', respectively. -------------------------------------------------------------------------------- -*/ -static void - normalizeFloat64Subnormal( bits64 aSig, int16 *zExpPtr, bits64 *zSigPtr ) -{ - int8 shiftCount; - - shiftCount = countLeadingZeros64( aSig ) - 11; - *zSigPtr = aSig<>10; - zSig &= ~ ( ( ( roundBits ^ 0x200 ) == 0 ) & roundNearestEven ); - if ( zSig == 0 ) zExp = 0; - return packFloat64( zSign, zExp, zSig ); - -} - -/* -------------------------------------------------------------------------------- -Takes an abstract floating-point value having sign `zSign', exponent `zExp', -and significand `zSig', and returns the proper double-precision floating- -point value corresponding to the abstract input. This routine is just like -`roundAndPackFloat64' except that `zSig' does not have to be normalized in -any way. In all cases, `zExp' must be 1 less than the ``true'' floating- -point exponent. -------------------------------------------------------------------------------- -*/ -static float64 - normalizeRoundAndPackFloat64( flag zSign, int16 zExp, bits64 zSig ) -{ - int8 shiftCount; - - shiftCount = countLeadingZeros64( zSig ) - 1; - return roundAndPackFloat64( zSign, zExp - shiftCount, zSig<>15; - -} - -/* -------------------------------------------------------------------------------- -Normalizes the subnormal extended double-precision floating-point value -represented by the denormalized significand `aSig'. The normalized exponent -and significand are stored at the locations pointed to by `zExpPtr' and -`zSigPtr', respectively. -------------------------------------------------------------------------------- -*/ -static void - normalizeFloatx80Subnormal( bits64 aSig, int32 *zExpPtr, bits64 *zSigPtr ) -{ - int8 shiftCount; - - shiftCount = countLeadingZeros64( aSig ); - *zSigPtr = aSig<>( - shiftCount ); - if ( (bits32) ( aSig<<( shiftCount & 31 ) ) ) { - float_exception_flags |= float_flag_inexact; - } - return aSign ? - z : z; - -} - -/* -------------------------------------------------------------------------------- -Returns the result of converting the single-precision floating-point value -`a' to the double-precision floating-point format. The conversion is -performed according to the IEC/IEEE Standard for Binary Floating-point -Arithmetic. -------------------------------------------------------------------------------- -*/ -float64 float32_to_float64( float32 a ) -{ - flag aSign; - int16 aExp; - bits32 aSig; - - aSig = extractFloat32Frac( a ); - aExp = extractFloat32Exp( a ); - aSign = extractFloat32Sign( a ); - if ( aExp == 0xFF ) { - if ( aSig ) return commonNaNToFloat64( float32ToCommonNaN( a ) ); - return packFloat64( aSign, 0x7FF, 0 ); - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return packFloat64( aSign, 0, 0 ); - normalizeFloat32Subnormal( aSig, &aExp, &aSig ); - --aExp; - } - return packFloat64( aSign, aExp + 0x380, ( (bits64) aSig )<<29 ); - -} - -#ifdef FLOATX80 - -/* -------------------------------------------------------------------------------- -Returns the result of converting the single-precision floating-point value -`a' to the extended double-precision floating-point format. The conversion -is performed according to the IEC/IEEE Standard for Binary Floating-point -Arithmetic. -------------------------------------------------------------------------------- -*/ -floatx80 float32_to_floatx80( float32 a ) -{ - flag aSign; - int16 aExp; - bits32 aSig; - - aSig = extractFloat32Frac( a ); - aExp = extractFloat32Exp( a ); - aSign = extractFloat32Sign( a ); - if ( aExp == 0xFF ) { - if ( aSig ) return commonNaNToFloatx80( float32ToCommonNaN( a ) ); - return packFloatx80( aSign, 0x7FFF, LIT64( 0x8000000000000000 ) ); - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return packFloatx80( aSign, 0, 0 ); - normalizeFloat32Subnormal( aSig, &aExp, &aSig ); - } - aSig |= 0x00800000; - return packFloatx80( aSign, aExp + 0x3F80, ( (bits64) aSig )<<40 ); - -} - -#endif - -/* -------------------------------------------------------------------------------- -Rounds the single-precision floating-point value `a' to an integer, and -returns the result as a single-precision floating-point value. The -operation is performed according to the IEC/IEEE Standard for Binary -Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float32 float32_round_to_int( float32 a ) -{ - flag aSign; - int16 aExp; - bits32 lastBitMask, roundBitsMask; - int8 roundingMode; - float32 z; - - aExp = extractFloat32Exp( a ); - if ( 0x96 <= aExp ) { - if ( ( aExp == 0xFF ) && extractFloat32Frac( a ) ) { - return propagateFloat32NaN( a, a ); - } - return a; - } - if ( aExp <= 0x7E ) { - if ( (bits32) ( a<<1 ) == 0 ) return a; - float_exception_flags |= float_flag_inexact; - aSign = extractFloat32Sign( a ); - switch ( float_rounding_mode ) { - case float_round_nearest_even: - if ( ( aExp == 0x7E ) && extractFloat32Frac( a ) ) { - return packFloat32( aSign, 0x7F, 0 ); - } - break; - case float_round_down: - return aSign ? 0xBF800000 : 0; - case float_round_up: - return aSign ? 0x80000000 : 0x3F800000; - } - return packFloat32( aSign, 0, 0 ); - } - lastBitMask = 1; - lastBitMask <<= 0x96 - aExp; - roundBitsMask = lastBitMask - 1; - z = a; - roundingMode = float_rounding_mode; - if ( roundingMode == float_round_nearest_even ) { - z += lastBitMask>>1; - if ( ( z & roundBitsMask ) == 0 ) z &= ~ lastBitMask; - } - else if ( roundingMode != float_round_to_zero ) { - if ( extractFloat32Sign( z ) ^ ( roundingMode == float_round_up ) ) { - z += roundBitsMask; - } - } - z &= ~ roundBitsMask; - if ( z != a ) float_exception_flags |= float_flag_inexact; - return z; - -} - -/* -------------------------------------------------------------------------------- -Returns the result of adding the absolute values of the single-precision -floating-point values `a' and `b'. If `zSign' is true, the sum is negated -before being returned. `zSign' is ignored if the result is a NaN. The -addition is performed according to the IEC/IEEE Standard for Binary -Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -static float32 addFloat32Sigs( float32 a, float32 b, flag zSign ) -{ - int16 aExp, bExp, zExp; - bits32 aSig, bSig, zSig; - int16 expDiff; - - aSig = extractFloat32Frac( a ); - aExp = extractFloat32Exp( a ); - bSig = extractFloat32Frac( b ); - bExp = extractFloat32Exp( b ); - expDiff = aExp - bExp; - aSig <<= 6; - bSig <<= 6; - if ( 0 < expDiff ) { - if ( aExp == 0xFF ) { - if ( aSig ) return propagateFloat32NaN( a, b ); - return a; - } - if ( bExp == 0 ) { - --expDiff; - } - else { - bSig |= 0x20000000; - } - shift32RightJamming( bSig, expDiff, &bSig ); - zExp = aExp; - } - else if ( expDiff < 0 ) { - if ( bExp == 0xFF ) { - if ( bSig ) return propagateFloat32NaN( a, b ); - return packFloat32( zSign, 0xFF, 0 ); - } - if ( aExp == 0 ) { - ++expDiff; - } - else { - aSig |= 0x20000000; - } - shift32RightJamming( aSig, - expDiff, &aSig ); - zExp = bExp; - } - else { - if ( aExp == 0xFF ) { - if ( aSig | bSig ) return propagateFloat32NaN( a, b ); - return a; - } - if ( aExp == 0 ) return packFloat32( zSign, 0, ( aSig + bSig )>>6 ); - zSig = 0x40000000 + aSig + bSig; - zExp = aExp; - goto roundAndPack; - } - aSig |= 0x20000000; - zSig = ( aSig + bSig )<<1; - --zExp; - if ( (sbits32) zSig < 0 ) { - zSig = aSig + bSig; - ++zExp; - } - roundAndPack: - return roundAndPackFloat32( zSign, zExp, zSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of subtracting the absolute values of the single- -precision floating-point values `a' and `b'. If `zSign' is true, the -difference is negated before being returned. `zSign' is ignored if the -result is a NaN. The subtraction is performed according to the IEC/IEEE -Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -static float32 subFloat32Sigs( float32 a, float32 b, flag zSign ) -{ - int16 aExp, bExp, zExp; - bits32 aSig, bSig, zSig; - int16 expDiff; - - aSig = extractFloat32Frac( a ); - aExp = extractFloat32Exp( a ); - bSig = extractFloat32Frac( b ); - bExp = extractFloat32Exp( b ); - expDiff = aExp - bExp; - aSig <<= 7; - bSig <<= 7; - if ( 0 < expDiff ) goto aExpBigger; - if ( expDiff < 0 ) goto bExpBigger; - if ( aExp == 0xFF ) { - if ( aSig | bSig ) return propagateFloat32NaN( a, b ); - float_raise( float_flag_invalid ); - return float32_default_nan; - } - if ( aExp == 0 ) { - aExp = 1; - bExp = 1; - } - if ( bSig < aSig ) goto aBigger; - if ( aSig < bSig ) goto bBigger; - return packFloat32( float_rounding_mode == float_round_down, 0, 0 ); - bExpBigger: - if ( bExp == 0xFF ) { - if ( bSig ) return propagateFloat32NaN( a, b ); - return packFloat32( zSign ^ 1, 0xFF, 0 ); - } - if ( aExp == 0 ) { - ++expDiff; - } - else { - aSig |= 0x40000000; - } - shift32RightJamming( aSig, - expDiff, &aSig ); - bSig |= 0x40000000; - bBigger: - zSig = bSig - aSig; - zExp = bExp; - zSign ^= 1; - goto normalizeRoundAndPack; - aExpBigger: - if ( aExp == 0xFF ) { - if ( aSig ) return propagateFloat32NaN( a, b ); - return a; - } - if ( bExp == 0 ) { - --expDiff; - } - else { - bSig |= 0x40000000; - } - shift32RightJamming( bSig, expDiff, &bSig ); - aSig |= 0x40000000; - aBigger: - zSig = aSig - bSig; - zExp = aExp; - normalizeRoundAndPack: - --zExp; - return normalizeRoundAndPackFloat32( zSign, zExp, zSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of adding the single-precision floating-point values `a' -and `b'. The operation is performed according to the IEC/IEEE Standard for -Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float32 float32_add( float32 a, float32 b ) -{ - flag aSign, bSign; - - aSign = extractFloat32Sign( a ); - bSign = extractFloat32Sign( b ); - if ( aSign == bSign ) { - return addFloat32Sigs( a, b, aSign ); - } - else { - return subFloat32Sigs( a, b, aSign ); - } - -} - -/* -------------------------------------------------------------------------------- -Returns the result of subtracting the single-precision floating-point values -`a' and `b'. The operation is performed according to the IEC/IEEE Standard -for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float32 float32_sub( float32 a, float32 b ) -{ - flag aSign, bSign; - - aSign = extractFloat32Sign( a ); - bSign = extractFloat32Sign( b ); - if ( aSign == bSign ) { - return subFloat32Sigs( a, b, aSign ); - } - else { - return addFloat32Sigs( a, b, aSign ); - } - -} - -/* -------------------------------------------------------------------------------- -Returns the result of multiplying the single-precision floating-point values -`a' and `b'. The operation is performed according to the IEC/IEEE Standard -for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float32 float32_mul( float32 a, float32 b ) -{ - flag aSign, bSign, zSign; - int16 aExp, bExp, zExp; - bits32 aSig, bSig; - bits64 zSig64; - bits32 zSig; - - aSig = extractFloat32Frac( a ); - aExp = extractFloat32Exp( a ); - aSign = extractFloat32Sign( a ); - bSig = extractFloat32Frac( b ); - bExp = extractFloat32Exp( b ); - bSign = extractFloat32Sign( b ); - zSign = aSign ^ bSign; - if ( aExp == 0xFF ) { - if ( aSig || ( ( bExp == 0xFF ) && bSig ) ) { - return propagateFloat32NaN( a, b ); - } - if ( ( bExp | bSig ) == 0 ) { - float_raise( float_flag_invalid ); - return float32_default_nan; - } - return packFloat32( zSign, 0xFF, 0 ); - } - if ( bExp == 0xFF ) { - if ( bSig ) return propagateFloat32NaN( a, b ); - if ( ( aExp | aSig ) == 0 ) { - float_raise( float_flag_invalid ); - return float32_default_nan; - } - return packFloat32( zSign, 0xFF, 0 ); - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return packFloat32( zSign, 0, 0 ); - normalizeFloat32Subnormal( aSig, &aExp, &aSig ); - } - if ( bExp == 0 ) { - if ( bSig == 0 ) return packFloat32( zSign, 0, 0 ); - normalizeFloat32Subnormal( bSig, &bExp, &bSig ); - } - zExp = aExp + bExp - 0x7F; - aSig = ( aSig | 0x00800000 )<<7; - bSig = ( bSig | 0x00800000 )<<8; - shift64RightJamming( ( (bits64) aSig ) * bSig, 32, &zSig64 ); - zSig = zSig64; - if ( 0 <= (sbits32) ( zSig<<1 ) ) { - zSig <<= 1; - --zExp; - } - return roundAndPackFloat32( zSign, zExp, zSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of dividing the single-precision floating-point value `a' -by the corresponding value `b'. The operation is performed according to the -IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float32 float32_div( float32 a, float32 b ) -{ - flag aSign, bSign, zSign; - int16 aExp, bExp, zExp; - bits32 aSig, bSig, zSig; - - aSig = extractFloat32Frac( a ); - aExp = extractFloat32Exp( a ); - aSign = extractFloat32Sign( a ); - bSig = extractFloat32Frac( b ); - bExp = extractFloat32Exp( b ); - bSign = extractFloat32Sign( b ); - zSign = aSign ^ bSign; - if ( aExp == 0xFF ) { - if ( aSig ) return propagateFloat32NaN( a, b ); - if ( bExp == 0xFF ) { - if ( bSig ) return propagateFloat32NaN( a, b ); - float_raise( float_flag_invalid ); - return float32_default_nan; - } - return packFloat32( zSign, 0xFF, 0 ); - } - if ( bExp == 0xFF ) { - if ( bSig ) return propagateFloat32NaN( a, b ); - return packFloat32( zSign, 0, 0 ); - } - if ( bExp == 0 ) { - if ( bSig == 0 ) { - if ( ( aExp | aSig ) == 0 ) { - float_raise( float_flag_invalid ); - return float32_default_nan; - } - float_raise( float_flag_divbyzero ); - return packFloat32( zSign, 0xFF, 0 ); - } - normalizeFloat32Subnormal( bSig, &bExp, &bSig ); - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return packFloat32( zSign, 0, 0 ); - normalizeFloat32Subnormal( aSig, &aExp, &aSig ); - } - zExp = aExp - bExp + 0x7D; - aSig = ( aSig | 0x00800000 )<<7; - bSig = ( bSig | 0x00800000 )<<8; - if ( bSig <= ( aSig + aSig ) ) { - aSig >>= 1; - ++zExp; - } - zSig = ( ( (bits64) aSig )<<32 ) / bSig; - if ( ( zSig & 0x3F ) == 0 ) { - zSig |= ( ( (bits64) bSig ) * zSig != ( (bits64) aSig )<<32 ); - } - return roundAndPackFloat32( zSign, zExp, zSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns the remainder of the single-precision floating-point value `a' -with respect to the corresponding value `b'. The operation is performed -according to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float32 float32_rem( float32 a, float32 b ) -{ - flag aSign, bSign, zSign; - int16 aExp, bExp, expDiff; - bits32 aSig, bSig; - bits32 q; - bits64 aSig64, bSig64, q64; - bits32 alternateASig; - sbits32 sigMean; - - aSig = extractFloat32Frac( a ); - aExp = extractFloat32Exp( a ); - aSign = extractFloat32Sign( a ); - bSig = extractFloat32Frac( b ); - bExp = extractFloat32Exp( b ); - bSign = extractFloat32Sign( b ); - if ( aExp == 0xFF ) { - if ( aSig || ( ( bExp == 0xFF ) && bSig ) ) { - return propagateFloat32NaN( a, b ); - } - float_raise( float_flag_invalid ); - return float32_default_nan; - } - if ( bExp == 0xFF ) { - if ( bSig ) return propagateFloat32NaN( a, b ); - return a; - } - if ( bExp == 0 ) { - if ( bSig == 0 ) { - float_raise( float_flag_invalid ); - return float32_default_nan; - } - normalizeFloat32Subnormal( bSig, &bExp, &bSig ); - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return a; - normalizeFloat32Subnormal( aSig, &aExp, &aSig ); - } - expDiff = aExp - bExp; - aSig |= 0x00800000; - bSig |= 0x00800000; - if ( expDiff < 32 ) { - aSig <<= 8; - bSig <<= 8; - if ( expDiff < 0 ) { - if ( expDiff < -1 ) return a; - aSig >>= 1; - } - q = ( bSig <= aSig ); - if ( q ) aSig -= bSig; - if ( 0 < expDiff ) { - q = ( ( (bits64) aSig )<<32 ) / bSig; - q >>= 32 - expDiff; - bSig >>= 2; - aSig = ( ( aSig>>1 )<<( expDiff - 1 ) ) - bSig * q; - } - else { - aSig >>= 2; - bSig >>= 2; - } - } - else { - if ( bSig <= aSig ) aSig -= bSig; - aSig64 = ( (bits64) aSig )<<40; - bSig64 = ( (bits64) bSig )<<40; - expDiff -= 64; - while ( 0 < expDiff ) { - q64 = estimateDiv128To64( aSig64, 0, bSig64 ); - q64 = ( 2 < q64 ) ? q64 - 2 : 0; - aSig64 = - ( ( bSig * q64 )<<38 ); - expDiff -= 62; - } - expDiff += 64; - q64 = estimateDiv128To64( aSig64, 0, bSig64 ); - q64 = ( 2 < q64 ) ? q64 - 2 : 0; - q = q64>>( 64 - expDiff ); - bSig <<= 6; - aSig = ( ( aSig64>>33 )<<( expDiff - 1 ) ) - bSig * q; - } - do { - alternateASig = aSig; - ++q; - aSig -= bSig; - } while ( 0 <= (sbits32) aSig ); - sigMean = aSig + alternateASig; - if ( ( sigMean < 0 ) || ( ( sigMean == 0 ) && ( q & 1 ) ) ) { - aSig = alternateASig; - } - zSign = ( (sbits32) aSig < 0 ); - if ( zSign ) aSig = - aSig; - return normalizeRoundAndPackFloat32( aSign ^ zSign, bExp, aSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns the square root of the single-precision floating-point value `a'. -The operation is performed according to the IEC/IEEE Standard for Binary -Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float32 float32_sqrt( float32 a ) -{ - flag aSign; - int16 aExp, zExp; - bits32 aSig, zSig; - bits64 rem, term; - - aSig = extractFloat32Frac( a ); - aExp = extractFloat32Exp( a ); - aSign = extractFloat32Sign( a ); - if ( aExp == 0xFF ) { - if ( aSig ) return propagateFloat32NaN( a, 0 ); - if ( ! aSign ) return a; - float_raise( float_flag_invalid ); - return float32_default_nan; - } - if ( aSign ) { - if ( ( aExp | aSig ) == 0 ) return a; - float_raise( float_flag_invalid ); - return float32_default_nan; - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return 0; - normalizeFloat32Subnormal( aSig, &aExp, &aSig ); - } - zExp = ( ( aExp - 0x7F )>>1 ) + 0x7E; - aSig = ( aSig | 0x00800000 )<<8; - zSig = estimateSqrt32( aExp, aSig ) + 2; - if ( ( zSig & 0x7F ) <= 5 ) { - if ( zSig < 2 ) { - zSig = 0xFFFFFFFF; - } - else { - aSig >>= aExp & 1; - term = ( (bits64) zSig ) * zSig; - rem = ( ( (bits64) aSig )<<32 ) - term; - while ( (sbits64) rem < 0 ) { - --zSig; - rem += ( ( (bits64) zSig )<<1 ) | 1; - } - zSig |= ( rem != 0 ); - } - } - shift32RightJamming( zSig, 1, &zSig ); - return roundAndPackFloat32( 0, zExp, zSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the single-precision floating-point value `a' is equal to the -corresponding value `b', and 0 otherwise. The comparison is performed -according to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float32_eq( float32 a, float32 b ) -{ - - if ( ( ( extractFloat32Exp( a ) == 0xFF ) && extractFloat32Frac( a ) ) - || ( ( extractFloat32Exp( b ) == 0xFF ) && extractFloat32Frac( b ) ) - ) { - if ( float32_is_signaling_nan( a ) || float32_is_signaling_nan( b ) ) { - float_raise( float_flag_invalid ); - } - return 0; - } - return ( a == b ) || ( (bits32) ( ( a | b )<<1 ) == 0 ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the single-precision floating-point value `a' is less than or -equal to the corresponding value `b', and 0 otherwise. The comparison is -performed according to the IEC/IEEE Standard for Binary Floating-point -Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float32_le( float32 a, float32 b ) -{ - flag aSign, bSign; - - if ( ( ( extractFloat32Exp( a ) == 0xFF ) && extractFloat32Frac( a ) ) - || ( ( extractFloat32Exp( b ) == 0xFF ) && extractFloat32Frac( b ) ) - ) { - float_raise( float_flag_invalid ); - return 0; - } - aSign = extractFloat32Sign( a ); - bSign = extractFloat32Sign( b ); - if ( aSign != bSign ) return aSign || ( (bits32) ( ( a | b )<<1 ) == 0 ); - return ( a == b ) || ( aSign ^ ( a < b ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the single-precision floating-point value `a' is less than -the corresponding value `b', and 0 otherwise. The comparison is performed -according to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float32_lt( float32 a, float32 b ) -{ - flag aSign, bSign; - - if ( ( ( extractFloat32Exp( a ) == 0xFF ) && extractFloat32Frac( a ) ) - || ( ( extractFloat32Exp( b ) == 0xFF ) && extractFloat32Frac( b ) ) - ) { - float_raise( float_flag_invalid ); - return 0; - } - aSign = extractFloat32Sign( a ); - bSign = extractFloat32Sign( b ); - if ( aSign != bSign ) return aSign && ( (bits32) ( ( a | b )<<1 ) != 0 ); - return ( a != b ) && ( aSign ^ ( a < b ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the single-precision floating-point value `a' is equal to the -corresponding value `b', and 0 otherwise. The invalid exception is raised -if either operand is a NaN. Otherwise, the comparison is performed -according to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float32_eq_signaling( float32 a, float32 b ) -{ - - if ( ( ( extractFloat32Exp( a ) == 0xFF ) && extractFloat32Frac( a ) ) - || ( ( extractFloat32Exp( b ) == 0xFF ) && extractFloat32Frac( b ) ) - ) { - float_raise( float_flag_invalid ); - return 0; - } - return ( a == b ) || ( (bits32) ( ( a | b )<<1 ) == 0 ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the single-precision floating-point value `a' is less than or -equal to the corresponding value `b', and 0 otherwise. Quiet NaNs do not -cause an exception. Otherwise, the comparison is performed according to the -IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float32_le_quiet( float32 a, float32 b ) -{ - flag aSign, bSign; - //int16 aExp, bExp; - - if ( ( ( extractFloat32Exp( a ) == 0xFF ) && extractFloat32Frac( a ) ) - || ( ( extractFloat32Exp( b ) == 0xFF ) && extractFloat32Frac( b ) ) - ) { - if ( float32_is_signaling_nan( a ) || float32_is_signaling_nan( b ) ) { - float_raise( float_flag_invalid ); - } - return 0; - } - aSign = extractFloat32Sign( a ); - bSign = extractFloat32Sign( b ); - if ( aSign != bSign ) return aSign || ( (bits32) ( ( a | b )<<1 ) == 0 ); - return ( a == b ) || ( aSign ^ ( a < b ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the single-precision floating-point value `a' is less than -the corresponding value `b', and 0 otherwise. Quiet NaNs do not cause an -exception. Otherwise, the comparison is performed according to the IEC/IEEE -Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float32_lt_quiet( float32 a, float32 b ) -{ - flag aSign, bSign; - - if ( ( ( extractFloat32Exp( a ) == 0xFF ) && extractFloat32Frac( a ) ) - || ( ( extractFloat32Exp( b ) == 0xFF ) && extractFloat32Frac( b ) ) - ) { - if ( float32_is_signaling_nan( a ) || float32_is_signaling_nan( b ) ) { - float_raise( float_flag_invalid ); - } - return 0; - } - aSign = extractFloat32Sign( a ); - bSign = extractFloat32Sign( b ); - if ( aSign != bSign ) return aSign && ( (bits32) ( ( a | b )<<1 ) != 0 ); - return ( a != b ) && ( aSign ^ ( a < b ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of converting the double-precision floating-point value -`a' to the 32-bit two's complement integer format. The conversion is -performed according to the IEC/IEEE Standard for Binary Floating-point -Arithmetic---which means in particular that the conversion is rounded -according to the current rounding mode. If `a' is a NaN, the largest -positive integer is returned. Otherwise, if the conversion overflows, the -largest integer with the same sign as `a' is returned. -------------------------------------------------------------------------------- -*/ -int32 float64_to_int32( float64 a ) -{ - flag aSign; - int16 aExp, shiftCount; - bits64 aSig; - - aSig = extractFloat64Frac( a ); - aExp = extractFloat64Exp( a ); - aSign = extractFloat64Sign( a ); - if ( ( aExp == 0x7FF ) && aSig ) aSign = 0; - if ( aExp ) aSig |= LIT64( 0x0010000000000000 ); - shiftCount = 0x42C - aExp; - if ( 0 < shiftCount ) shift64RightJamming( aSig, shiftCount, &aSig ); - return roundAndPackInt32( aSign, aSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of converting the double-precision floating-point value -`a' to the 32-bit two's complement integer format. The conversion is -performed according to the IEC/IEEE Standard for Binary Floating-point -Arithmetic, except that the conversion is always rounded toward zero. If -`a' is a NaN, the largest positive integer is returned. Otherwise, if the -conversion overflows, the largest integer with the same sign as `a' is -returned. -------------------------------------------------------------------------------- -*/ -int32 float64_to_int32_round_to_zero( float64 a ) -{ - flag aSign; - int16 aExp, shiftCount; - bits64 aSig, savedASig; - int32 z; - - aSig = extractFloat64Frac( a ); - aExp = extractFloat64Exp( a ); - aSign = extractFloat64Sign( a ); - shiftCount = 0x433 - aExp; - if ( shiftCount < 21 ) { - if ( ( aExp == 0x7FF ) && aSig ) aSign = 0; - goto invalid; - } - else if ( 52 < shiftCount ) { - if ( aExp || aSig ) float_exception_flags |= float_flag_inexact; - return 0; - } - aSig |= LIT64( 0x0010000000000000 ); - savedASig = aSig; - aSig >>= shiftCount; - z = aSig; - if ( aSign ) z = - z; - if ( ( z < 0 ) ^ aSign ) { - invalid: - float_exception_flags |= float_flag_invalid; - return aSign ? 0x80000000 : 0x7FFFFFFF; - } - if ( ( aSig<>= shiftCount; - z = aSig; - if ( aSign ) z = - z; - if ( ( z < 0 ) ^ aSign ) { - invalid: - float_exception_flags |= float_flag_invalid; - return aSign ? 0x80000000 : 0x7FFFFFFF; - } - if ( ( aSig<>1; - if ( ( z & roundBitsMask ) == 0 ) z &= ~ lastBitMask; - } - else if ( roundingMode != float_round_to_zero ) { - if ( extractFloat64Sign( z ) ^ ( roundingMode == float_round_up ) ) { - z += roundBitsMask; - } - } - z &= ~ roundBitsMask; - if ( z != a ) float_exception_flags |= float_flag_inexact; - return z; - -} - -/* -------------------------------------------------------------------------------- -Returns the result of adding the absolute values of the double-precision -floating-point values `a' and `b'. If `zSign' is true, the sum is negated -before being returned. `zSign' is ignored if the result is a NaN. The -addition is performed according to the IEC/IEEE Standard for Binary -Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -static float64 addFloat64Sigs( float64 a, float64 b, flag zSign ) -{ - int16 aExp, bExp, zExp; - bits64 aSig, bSig, zSig; - int16 expDiff; - - aSig = extractFloat64Frac( a ); - aExp = extractFloat64Exp( a ); - bSig = extractFloat64Frac( b ); - bExp = extractFloat64Exp( b ); - expDiff = aExp - bExp; - aSig <<= 9; - bSig <<= 9; - if ( 0 < expDiff ) { - if ( aExp == 0x7FF ) { - if ( aSig ) return propagateFloat64NaN( a, b ); - return a; - } - if ( bExp == 0 ) { - --expDiff; - } - else { - bSig |= LIT64( 0x2000000000000000 ); - } - shift64RightJamming( bSig, expDiff, &bSig ); - zExp = aExp; - } - else if ( expDiff < 0 ) { - if ( bExp == 0x7FF ) { - if ( bSig ) return propagateFloat64NaN( a, b ); - return packFloat64( zSign, 0x7FF, 0 ); - } - if ( aExp == 0 ) { - ++expDiff; - } - else { - aSig |= LIT64( 0x2000000000000000 ); - } - shift64RightJamming( aSig, - expDiff, &aSig ); - zExp = bExp; - } - else { - if ( aExp == 0x7FF ) { - if ( aSig | bSig ) return propagateFloat64NaN( a, b ); - return a; - } - if ( aExp == 0 ) return packFloat64( zSign, 0, ( aSig + bSig )>>9 ); - zSig = LIT64( 0x4000000000000000 ) + aSig + bSig; - zExp = aExp; - goto roundAndPack; - } - aSig |= LIT64( 0x2000000000000000 ); - zSig = ( aSig + bSig )<<1; - --zExp; - if ( (sbits64) zSig < 0 ) { - zSig = aSig + bSig; - ++zExp; - } - roundAndPack: - return roundAndPackFloat64( zSign, zExp, zSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of subtracting the absolute values of the double- -precision floating-point values `a' and `b'. If `zSign' is true, the -difference is negated before being returned. `zSign' is ignored if the -result is a NaN. The subtraction is performed according to the IEC/IEEE -Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -static float64 subFloat64Sigs( float64 a, float64 b, flag zSign ) -{ - int16 aExp, bExp, zExp; - bits64 aSig, bSig, zSig; - int16 expDiff; - - aSig = extractFloat64Frac( a ); - aExp = extractFloat64Exp( a ); - bSig = extractFloat64Frac( b ); - bExp = extractFloat64Exp( b ); - expDiff = aExp - bExp; - aSig <<= 10; - bSig <<= 10; - if ( 0 < expDiff ) goto aExpBigger; - if ( expDiff < 0 ) goto bExpBigger; - if ( aExp == 0x7FF ) { - if ( aSig | bSig ) return propagateFloat64NaN( a, b ); - float_raise( float_flag_invalid ); - return float64_default_nan; - } - if ( aExp == 0 ) { - aExp = 1; - bExp = 1; - } - if ( bSig < aSig ) goto aBigger; - if ( aSig < bSig ) goto bBigger; - return packFloat64( float_rounding_mode == float_round_down, 0, 0 ); - bExpBigger: - if ( bExp == 0x7FF ) { - if ( bSig ) return propagateFloat64NaN( a, b ); - return packFloat64( zSign ^ 1, 0x7FF, 0 ); - } - if ( aExp == 0 ) { - ++expDiff; - } - else { - aSig |= LIT64( 0x4000000000000000 ); - } - shift64RightJamming( aSig, - expDiff, &aSig ); - bSig |= LIT64( 0x4000000000000000 ); - bBigger: - zSig = bSig - aSig; - zExp = bExp; - zSign ^= 1; - goto normalizeRoundAndPack; - aExpBigger: - if ( aExp == 0x7FF ) { - if ( aSig ) return propagateFloat64NaN( a, b ); - return a; - } - if ( bExp == 0 ) { - --expDiff; - } - else { - bSig |= LIT64( 0x4000000000000000 ); - } - shift64RightJamming( bSig, expDiff, &bSig ); - aSig |= LIT64( 0x4000000000000000 ); - aBigger: - zSig = aSig - bSig; - zExp = aExp; - normalizeRoundAndPack: - --zExp; - return normalizeRoundAndPackFloat64( zSign, zExp, zSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of adding the double-precision floating-point values `a' -and `b'. The operation is performed according to the IEC/IEEE Standard for -Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float64 float64_add( float64 a, float64 b ) -{ - flag aSign, bSign; - - aSign = extractFloat64Sign( a ); - bSign = extractFloat64Sign( b ); - if ( aSign == bSign ) { - return addFloat64Sigs( a, b, aSign ); - } - else { - return subFloat64Sigs( a, b, aSign ); - } - -} - -/* -------------------------------------------------------------------------------- -Returns the result of subtracting the double-precision floating-point values -`a' and `b'. The operation is performed according to the IEC/IEEE Standard -for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float64 float64_sub( float64 a, float64 b ) -{ - flag aSign, bSign; - - aSign = extractFloat64Sign( a ); - bSign = extractFloat64Sign( b ); - if ( aSign == bSign ) { - return subFloat64Sigs( a, b, aSign ); - } - else { - return addFloat64Sigs( a, b, aSign ); - } - -} - -/* -------------------------------------------------------------------------------- -Returns the result of multiplying the double-precision floating-point values -`a' and `b'. The operation is performed according to the IEC/IEEE Standard -for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float64 float64_mul( float64 a, float64 b ) -{ - flag aSign, bSign, zSign; - int16 aExp, bExp, zExp; - bits64 aSig, bSig, zSig0, zSig1; - - aSig = extractFloat64Frac( a ); - aExp = extractFloat64Exp( a ); - aSign = extractFloat64Sign( a ); - bSig = extractFloat64Frac( b ); - bExp = extractFloat64Exp( b ); - bSign = extractFloat64Sign( b ); - zSign = aSign ^ bSign; - if ( aExp == 0x7FF ) { - if ( aSig || ( ( bExp == 0x7FF ) && bSig ) ) { - return propagateFloat64NaN( a, b ); - } - if ( ( bExp | bSig ) == 0 ) { - float_raise( float_flag_invalid ); - return float64_default_nan; - } - return packFloat64( zSign, 0x7FF, 0 ); - } - if ( bExp == 0x7FF ) { - if ( bSig ) return propagateFloat64NaN( a, b ); - if ( ( aExp | aSig ) == 0 ) { - float_raise( float_flag_invalid ); - return float64_default_nan; - } - return packFloat64( zSign, 0x7FF, 0 ); - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return packFloat64( zSign, 0, 0 ); - normalizeFloat64Subnormal( aSig, &aExp, &aSig ); - } - if ( bExp == 0 ) { - if ( bSig == 0 ) return packFloat64( zSign, 0, 0 ); - normalizeFloat64Subnormal( bSig, &bExp, &bSig ); - } - zExp = aExp + bExp - 0x3FF; - aSig = ( aSig | LIT64( 0x0010000000000000 ) )<<10; - bSig = ( bSig | LIT64( 0x0010000000000000 ) )<<11; - mul64To128( aSig, bSig, &zSig0, &zSig1 ); - zSig0 |= ( zSig1 != 0 ); - if ( 0 <= (sbits64) ( zSig0<<1 ) ) { - zSig0 <<= 1; - --zExp; - } - return roundAndPackFloat64( zSign, zExp, zSig0 ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of dividing the double-precision floating-point value `a' -by the corresponding value `b'. The operation is performed according to -the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float64 float64_div( float64 a, float64 b ) -{ - flag aSign, bSign, zSign; - int16 aExp, bExp, zExp; - bits64 aSig, bSig, zSig; - bits64 rem0, rem1; - bits64 term0, term1; - - aSig = extractFloat64Frac( a ); - aExp = extractFloat64Exp( a ); - aSign = extractFloat64Sign( a ); - bSig = extractFloat64Frac( b ); - bExp = extractFloat64Exp( b ); - bSign = extractFloat64Sign( b ); - zSign = aSign ^ bSign; - if ( aExp == 0x7FF ) { - if ( aSig ) return propagateFloat64NaN( a, b ); - if ( bExp == 0x7FF ) { - if ( bSig ) return propagateFloat64NaN( a, b ); - float_raise( float_flag_invalid ); - return float64_default_nan; - } - return packFloat64( zSign, 0x7FF, 0 ); - } - if ( bExp == 0x7FF ) { - if ( bSig ) return propagateFloat64NaN( a, b ); - return packFloat64( zSign, 0, 0 ); - } - if ( bExp == 0 ) { - if ( bSig == 0 ) { - if ( ( aExp | aSig ) == 0 ) { - float_raise( float_flag_invalid ); - return float64_default_nan; - } - float_raise( float_flag_divbyzero ); - return packFloat64( zSign, 0x7FF, 0 ); - } - normalizeFloat64Subnormal( bSig, &bExp, &bSig ); - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return packFloat64( zSign, 0, 0 ); - normalizeFloat64Subnormal( aSig, &aExp, &aSig ); - } - zExp = aExp - bExp + 0x3FD; - aSig = ( aSig | LIT64( 0x0010000000000000 ) )<<10; - bSig = ( bSig | LIT64( 0x0010000000000000 ) )<<11; - if ( bSig <= ( aSig + aSig ) ) { - aSig >>= 1; - ++zExp; - } - zSig = estimateDiv128To64( aSig, 0, bSig ); - if ( ( zSig & 0x1FF ) <= 2 ) { - mul64To128( bSig, zSig, &term0, &term1 ); - sub128( aSig, 0, term0, term1, &rem0, &rem1 ); - while ( (sbits64) rem0 < 0 ) { - --zSig; - add128( rem0, rem1, 0, bSig, &rem0, &rem1 ); - } - zSig |= ( rem1 != 0 ); - } - return roundAndPackFloat64( zSign, zExp, zSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns the remainder of the double-precision floating-point value `a' -with respect to the corresponding value `b'. The operation is performed -according to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float64 float64_rem( float64 a, float64 b ) -{ - flag aSign, bSign, zSign; - int16 aExp, bExp, expDiff; - bits64 aSig, bSig; - bits64 q, alternateASig; - sbits64 sigMean; - - aSig = extractFloat64Frac( a ); - aExp = extractFloat64Exp( a ); - aSign = extractFloat64Sign( a ); - bSig = extractFloat64Frac( b ); - bExp = extractFloat64Exp( b ); - bSign = extractFloat64Sign( b ); - if ( aExp == 0x7FF ) { - if ( aSig || ( ( bExp == 0x7FF ) && bSig ) ) { - return propagateFloat64NaN( a, b ); - } - float_raise( float_flag_invalid ); - return float64_default_nan; - } - if ( bExp == 0x7FF ) { - if ( bSig ) return propagateFloat64NaN( a, b ); - return a; - } - if ( bExp == 0 ) { - if ( bSig == 0 ) { - float_raise( float_flag_invalid ); - return float64_default_nan; - } - normalizeFloat64Subnormal( bSig, &bExp, &bSig ); - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return a; - normalizeFloat64Subnormal( aSig, &aExp, &aSig ); - } - expDiff = aExp - bExp; - aSig = ( aSig | LIT64( 0x0010000000000000 ) )<<11; - bSig = ( bSig | LIT64( 0x0010000000000000 ) )<<11; - if ( expDiff < 0 ) { - if ( expDiff < -1 ) return a; - aSig >>= 1; - } - q = ( bSig <= aSig ); - if ( q ) aSig -= bSig; - expDiff -= 64; - while ( 0 < expDiff ) { - q = estimateDiv128To64( aSig, 0, bSig ); - q = ( 2 < q ) ? q - 2 : 0; - aSig = - ( ( bSig>>2 ) * q ); - expDiff -= 62; - } - expDiff += 64; - if ( 0 < expDiff ) { - q = estimateDiv128To64( aSig, 0, bSig ); - q = ( 2 < q ) ? q - 2 : 0; - q >>= 64 - expDiff; - bSig >>= 2; - aSig = ( ( aSig>>1 )<<( expDiff - 1 ) ) - bSig * q; - } - else { - aSig >>= 2; - bSig >>= 2; - } - do { - alternateASig = aSig; - ++q; - aSig -= bSig; - } while ( 0 <= (sbits64) aSig ); - sigMean = aSig + alternateASig; - if ( ( sigMean < 0 ) || ( ( sigMean == 0 ) && ( q & 1 ) ) ) { - aSig = alternateASig; - } - zSign = ( (sbits64) aSig < 0 ); - if ( zSign ) aSig = - aSig; - return normalizeRoundAndPackFloat64( aSign ^ zSign, bExp, aSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns the square root of the double-precision floating-point value `a'. -The operation is performed according to the IEC/IEEE Standard for Binary -Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -float64 float64_sqrt( float64 a ) -{ - flag aSign; - int16 aExp, zExp; - bits64 aSig, zSig; - bits64 rem0, rem1, term0, term1; //, shiftedRem; - //float64 z; - - aSig = extractFloat64Frac( a ); - aExp = extractFloat64Exp( a ); - aSign = extractFloat64Sign( a ); - if ( aExp == 0x7FF ) { - if ( aSig ) return propagateFloat64NaN( a, a ); - if ( ! aSign ) return a; - float_raise( float_flag_invalid ); - return float64_default_nan; - } - if ( aSign ) { - if ( ( aExp | aSig ) == 0 ) return a; - float_raise( float_flag_invalid ); - return float64_default_nan; - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return 0; - normalizeFloat64Subnormal( aSig, &aExp, &aSig ); - } - zExp = ( ( aExp - 0x3FF )>>1 ) + 0x3FE; - aSig |= LIT64( 0x0010000000000000 ); - zSig = estimateSqrt32( aExp, aSig>>21 ); - zSig <<= 31; - aSig <<= 9 - ( aExp & 1 ); - zSig = estimateDiv128To64( aSig, 0, zSig ) + zSig + 2; - if ( ( zSig & 0x3FF ) <= 5 ) { - if ( zSig < 2 ) { - zSig = LIT64( 0xFFFFFFFFFFFFFFFF ); - } - else { - aSig <<= 2; - mul64To128( zSig, zSig, &term0, &term1 ); - sub128( aSig, 0, term0, term1, &rem0, &rem1 ); - while ( (sbits64) rem0 < 0 ) { - --zSig; - shortShift128Left( 0, zSig, 1, &term0, &term1 ); - term1 |= 1; - add128( rem0, rem1, term0, term1, &rem0, &rem1 ); - } - zSig |= ( ( rem0 | rem1 ) != 0 ); - } - } - shift64RightJamming( zSig, 1, &zSig ); - return roundAndPackFloat64( 0, zExp, zSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the double-precision floating-point value `a' is equal to the -corresponding value `b', and 0 otherwise. The comparison is performed -according to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float64_eq( float64 a, float64 b ) -{ - - if ( ( ( extractFloat64Exp( a ) == 0x7FF ) && extractFloat64Frac( a ) ) - || ( ( extractFloat64Exp( b ) == 0x7FF ) && extractFloat64Frac( b ) ) - ) { - if ( float64_is_signaling_nan( a ) || float64_is_signaling_nan( b ) ) { - float_raise( float_flag_invalid ); - } - return 0; - } - return ( a == b ) || ( (bits64) ( ( a | b )<<1 ) == 0 ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the double-precision floating-point value `a' is less than or -equal to the corresponding value `b', and 0 otherwise. The comparison is -performed according to the IEC/IEEE Standard for Binary Floating-point -Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float64_le( float64 a, float64 b ) -{ - flag aSign, bSign; - - if ( ( ( extractFloat64Exp( a ) == 0x7FF ) && extractFloat64Frac( a ) ) - || ( ( extractFloat64Exp( b ) == 0x7FF ) && extractFloat64Frac( b ) ) - ) { - float_raise( float_flag_invalid ); - return 0; - } - aSign = extractFloat64Sign( a ); - bSign = extractFloat64Sign( b ); - if ( aSign != bSign ) return aSign || ( (bits64) ( ( a | b )<<1 ) == 0 ); - return ( a == b ) || ( aSign ^ ( a < b ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the double-precision floating-point value `a' is less than -the corresponding value `b', and 0 otherwise. The comparison is performed -according to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float64_lt( float64 a, float64 b ) -{ - flag aSign, bSign; - - if ( ( ( extractFloat64Exp( a ) == 0x7FF ) && extractFloat64Frac( a ) ) - || ( ( extractFloat64Exp( b ) == 0x7FF ) && extractFloat64Frac( b ) ) - ) { - float_raise( float_flag_invalid ); - return 0; - } - aSign = extractFloat64Sign( a ); - bSign = extractFloat64Sign( b ); - if ( aSign != bSign ) return aSign && ( (bits64) ( ( a | b )<<1 ) != 0 ); - return ( a != b ) && ( aSign ^ ( a < b ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the double-precision floating-point value `a' is equal to the -corresponding value `b', and 0 otherwise. The invalid exception is raised -if either operand is a NaN. Otherwise, the comparison is performed -according to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float64_eq_signaling( float64 a, float64 b ) -{ - - if ( ( ( extractFloat64Exp( a ) == 0x7FF ) && extractFloat64Frac( a ) ) - || ( ( extractFloat64Exp( b ) == 0x7FF ) && extractFloat64Frac( b ) ) - ) { - float_raise( float_flag_invalid ); - return 0; - } - return ( a == b ) || ( (bits64) ( ( a | b )<<1 ) == 0 ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the double-precision floating-point value `a' is less than or -equal to the corresponding value `b', and 0 otherwise. Quiet NaNs do not -cause an exception. Otherwise, the comparison is performed according to the -IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float64_le_quiet( float64 a, float64 b ) -{ - flag aSign, bSign; - //int16 aExp, bExp; - - if ( ( ( extractFloat64Exp( a ) == 0x7FF ) && extractFloat64Frac( a ) ) - || ( ( extractFloat64Exp( b ) == 0x7FF ) && extractFloat64Frac( b ) ) - ) { - if ( float64_is_signaling_nan( a ) || float64_is_signaling_nan( b ) ) { - float_raise( float_flag_invalid ); - } - return 0; - } - aSign = extractFloat64Sign( a ); - bSign = extractFloat64Sign( b ); - if ( aSign != bSign ) return aSign || ( (bits64) ( ( a | b )<<1 ) == 0 ); - return ( a == b ) || ( aSign ^ ( a < b ) ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the double-precision floating-point value `a' is less than -the corresponding value `b', and 0 otherwise. Quiet NaNs do not cause an -exception. Otherwise, the comparison is performed according to the IEC/IEEE -Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag float64_lt_quiet( float64 a, float64 b ) -{ - flag aSign, bSign; - - if ( ( ( extractFloat64Exp( a ) == 0x7FF ) && extractFloat64Frac( a ) ) - || ( ( extractFloat64Exp( b ) == 0x7FF ) && extractFloat64Frac( b ) ) - ) { - if ( float64_is_signaling_nan( a ) || float64_is_signaling_nan( b ) ) { - float_raise( float_flag_invalid ); - } - return 0; - } - aSign = extractFloat64Sign( a ); - bSign = extractFloat64Sign( b ); - if ( aSign != bSign ) return aSign && ( (bits64) ( ( a | b )<<1 ) != 0 ); - return ( a != b ) && ( aSign ^ ( a < b ) ); - -} - -#ifdef FLOATX80 - -/* -------------------------------------------------------------------------------- -Returns the result of converting the extended double-precision floating- -point value `a' to the 32-bit two's complement integer format. The -conversion is performed according to the IEC/IEEE Standard for Binary -Floating-point Arithmetic---which means in particular that the conversion -is rounded according to the current rounding mode. If `a' is a NaN, the -largest positive integer is returned. Otherwise, if the conversion -overflows, the largest integer with the same sign as `a' is returned. -------------------------------------------------------------------------------- -*/ -int32 floatx80_to_int32( floatx80 a ) -{ - flag aSign; - int32 aExp, shiftCount; - bits64 aSig; - - aSig = extractFloatx80Frac( a ); - aExp = extractFloatx80Exp( a ); - aSign = extractFloatx80Sign( a ); - if ( ( aExp == 0x7FFF ) && (bits64) ( aSig<<1 ) ) aSign = 0; - shiftCount = 0x4037 - aExp; - if ( shiftCount <= 0 ) shiftCount = 1; - shift64RightJamming( aSig, shiftCount, &aSig ); - return roundAndPackInt32( aSign, aSig ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of converting the extended double-precision floating- -point value `a' to the 32-bit two's complement integer format. The -conversion is performed according to the IEC/IEEE Standard for Binary -Floating-point Arithmetic, except that the conversion is always rounded -toward zero. If `a' is a NaN, the largest positive integer is returned. -Otherwise, if the conversion overflows, the largest integer with the same -sign as `a' is returned. -------------------------------------------------------------------------------- -*/ -int32 floatx80_to_int32_round_to_zero( floatx80 a ) -{ - flag aSign; - int32 aExp, shiftCount; - bits64 aSig, savedASig; - int32 z; - - aSig = extractFloatx80Frac( a ); - aExp = extractFloatx80Exp( a ); - aSign = extractFloatx80Sign( a ); - shiftCount = 0x403E - aExp; - if ( shiftCount < 32 ) { - if ( ( aExp == 0x7FFF ) && (bits64) ( aSig<<1 ) ) aSign = 0; - goto invalid; - } - else if ( 63 < shiftCount ) { - if ( aExp || aSig ) float_exception_flags |= float_flag_inexact; - return 0; - } - savedASig = aSig; - aSig >>= shiftCount; - z = aSig; - if ( aSign ) z = - z; - if ( ( z < 0 ) ^ aSign ) { - invalid: - float_exception_flags |= float_flag_invalid; - return aSign ? 0x80000000 : 0x7FFFFFFF; - } - if ( ( aSig<>1; - if ( ( z.low & roundBitsMask ) == 0 ) z.low &= ~ lastBitMask; - } - else if ( roundingMode != float_round_to_zero ) { - if ( extractFloatx80Sign( z ) ^ ( roundingMode == float_round_up ) ) { - z.low += roundBitsMask; - } - } - z.low &= ~ roundBitsMask; - if ( z.low == 0 ) { - ++z.high; - z.low = LIT64( 0x8000000000000000 ); - } - if ( z.low != a.low ) float_exception_flags |= float_flag_inexact; - return z; - -} - -/* -------------------------------------------------------------------------------- -Returns the result of adding the absolute values of the extended double- -precision floating-point values `a' and `b'. If `zSign' is true, the sum is -negated before being returned. `zSign' is ignored if the result is a NaN. -The addition is performed according to the IEC/IEEE Standard for Binary -Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -static floatx80 addFloatx80Sigs( floatx80 a, floatx80 b, flag zSign ) -{ - int32 aExp, bExp, zExp; - bits64 aSig, bSig, zSig0, zSig1; - int32 expDiff; - - aSig = extractFloatx80Frac( a ); - aExp = extractFloatx80Exp( a ); - bSig = extractFloatx80Frac( b ); - bExp = extractFloatx80Exp( b ); - expDiff = aExp - bExp; - if ( 0 < expDiff ) { - if ( aExp == 0x7FFF ) { - if ( (bits64) ( aSig<<1 ) ) return propagateFloatx80NaN( a, b ); - return a; - } - if ( bExp == 0 ) --expDiff; - shift64ExtraRightJamming( bSig, 0, expDiff, &bSig, &zSig1 ); - zExp = aExp; - } - else if ( expDiff < 0 ) { - if ( bExp == 0x7FFF ) { - if ( (bits64) ( bSig<<1 ) ) return propagateFloatx80NaN( a, b ); - return packFloatx80( zSign, 0x7FFF, LIT64( 0x8000000000000000 ) ); - } - if ( aExp == 0 ) ++expDiff; - shift64ExtraRightJamming( aSig, 0, - expDiff, &aSig, &zSig1 ); - zExp = bExp; - } - else { - if ( aExp == 0x7FFF ) { - if ( (bits64) ( ( aSig | bSig )<<1 ) ) { - return propagateFloatx80NaN( a, b ); - } - return a; - } - zSig1 = 0; - zSig0 = aSig + bSig; - if ( aExp == 0 ) { - normalizeFloatx80Subnormal( zSig0, &zExp, &zSig0 ); - goto roundAndPack; - } - zExp = aExp; - goto shiftRight1; - } - - zSig0 = aSig + bSig; - - if ( (sbits64) zSig0 < 0 ) goto roundAndPack; - shiftRight1: - shift64ExtraRightJamming( zSig0, zSig1, 1, &zSig0, &zSig1 ); - zSig0 |= LIT64( 0x8000000000000000 ); - ++zExp; - roundAndPack: - return - roundAndPackFloatx80( - floatx80_rounding_precision, zSign, zExp, zSig0, zSig1 ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of subtracting the absolute values of the extended -double-precision floating-point values `a' and `b'. If `zSign' is true, -the difference is negated before being returned. `zSign' is ignored if the -result is a NaN. The subtraction is performed according to the IEC/IEEE -Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -static floatx80 subFloatx80Sigs( floatx80 a, floatx80 b, flag zSign ) -{ - int32 aExp, bExp, zExp; - bits64 aSig, bSig, zSig0, zSig1; - int32 expDiff; - floatx80 z; - - aSig = extractFloatx80Frac( a ); - aExp = extractFloatx80Exp( a ); - bSig = extractFloatx80Frac( b ); - bExp = extractFloatx80Exp( b ); - expDiff = aExp - bExp; - if ( 0 < expDiff ) goto aExpBigger; - if ( expDiff < 0 ) goto bExpBigger; - if ( aExp == 0x7FFF ) { - if ( (bits64) ( ( aSig | bSig )<<1 ) ) { - return propagateFloatx80NaN( a, b ); - } - float_raise( float_flag_invalid ); - z.low = floatx80_default_nan_low; - z.high = floatx80_default_nan_high; - return z; - } - if ( aExp == 0 ) { - aExp = 1; - bExp = 1; - } - zSig1 = 0; - if ( bSig < aSig ) goto aBigger; - if ( aSig < bSig ) goto bBigger; - return packFloatx80( float_rounding_mode == float_round_down, 0, 0 ); - bExpBigger: - if ( bExp == 0x7FFF ) { - if ( (bits64) ( bSig<<1 ) ) return propagateFloatx80NaN( a, b ); - return packFloatx80( zSign ^ 1, 0x7FFF, LIT64( 0x8000000000000000 ) ); - } - if ( aExp == 0 ) ++expDiff; - shift128RightJamming( aSig, 0, - expDiff, &aSig, &zSig1 ); - bBigger: - sub128( bSig, 0, aSig, zSig1, &zSig0, &zSig1 ); - zExp = bExp; - zSign ^= 1; - goto normalizeRoundAndPack; - aExpBigger: - if ( aExp == 0x7FFF ) { - if ( (bits64) ( aSig<<1 ) ) return propagateFloatx80NaN( a, b ); - return a; - } - if ( bExp == 0 ) --expDiff; - shift128RightJamming( bSig, 0, expDiff, &bSig, &zSig1 ); - aBigger: - sub128( aSig, 0, bSig, zSig1, &zSig0, &zSig1 ); - zExp = aExp; - normalizeRoundAndPack: - return - normalizeRoundAndPackFloatx80( - floatx80_rounding_precision, zSign, zExp, zSig0, zSig1 ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of adding the extended double-precision floating-point -values `a' and `b'. The operation is performed according to the IEC/IEEE -Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -floatx80 floatx80_add( floatx80 a, floatx80 b ) -{ - flag aSign, bSign; - - aSign = extractFloatx80Sign( a ); - bSign = extractFloatx80Sign( b ); - if ( aSign == bSign ) { - return addFloatx80Sigs( a, b, aSign ); - } - else { - return subFloatx80Sigs( a, b, aSign ); - } - -} - -/* -------------------------------------------------------------------------------- -Returns the result of subtracting the extended double-precision floating- -point values `a' and `b'. The operation is performed according to the -IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -floatx80 floatx80_sub( floatx80 a, floatx80 b ) -{ - flag aSign, bSign; - - aSign = extractFloatx80Sign( a ); - bSign = extractFloatx80Sign( b ); - if ( aSign == bSign ) { - return subFloatx80Sigs( a, b, aSign ); - } - else { - return addFloatx80Sigs( a, b, aSign ); - } - -} - -/* -------------------------------------------------------------------------------- -Returns the result of multiplying the extended double-precision floating- -point values `a' and `b'. The operation is performed according to the -IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -floatx80 floatx80_mul( floatx80 a, floatx80 b ) -{ - flag aSign, bSign, zSign; - int32 aExp, bExp, zExp; - bits64 aSig, bSig, zSig0, zSig1; - floatx80 z; - - aSig = extractFloatx80Frac( a ); - aExp = extractFloatx80Exp( a ); - aSign = extractFloatx80Sign( a ); - bSig = extractFloatx80Frac( b ); - bExp = extractFloatx80Exp( b ); - bSign = extractFloatx80Sign( b ); - zSign = aSign ^ bSign; - if ( aExp == 0x7FFF ) { - if ( (bits64) ( aSig<<1 ) - || ( ( bExp == 0x7FFF ) && (bits64) ( bSig<<1 ) ) ) { - return propagateFloatx80NaN( a, b ); - } - if ( ( bExp | bSig ) == 0 ) goto invalid; - return packFloatx80( zSign, 0x7FFF, LIT64( 0x8000000000000000 ) ); - } - if ( bExp == 0x7FFF ) { - if ( (bits64) ( bSig<<1 ) ) return propagateFloatx80NaN( a, b ); - if ( ( aExp | aSig ) == 0 ) { - invalid: - float_raise( float_flag_invalid ); - z.low = floatx80_default_nan_low; - z.high = floatx80_default_nan_high; - return z; - } - return packFloatx80( zSign, 0x7FFF, LIT64( 0x8000000000000000 ) ); - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return packFloatx80( zSign, 0, 0 ); - normalizeFloatx80Subnormal( aSig, &aExp, &aSig ); - } - if ( bExp == 0 ) { - if ( bSig == 0 ) return packFloatx80( zSign, 0, 0 ); - normalizeFloatx80Subnormal( bSig, &bExp, &bSig ); - } - zExp = aExp + bExp - 0x3FFE; - mul64To128( aSig, bSig, &zSig0, &zSig1 ); - if ( 0 < (sbits64) zSig0 ) { - shortShift128Left( zSig0, zSig1, 1, &zSig0, &zSig1 ); - --zExp; - } - return - roundAndPackFloatx80( - floatx80_rounding_precision, zSign, zExp, zSig0, zSig1 ); - -} - -/* -------------------------------------------------------------------------------- -Returns the result of dividing the extended double-precision floating-point -value `a' by the corresponding value `b'. The operation is performed -according to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -floatx80 floatx80_div( floatx80 a, floatx80 b ) -{ - flag aSign, bSign, zSign; - int32 aExp, bExp, zExp; - bits64 aSig, bSig, zSig0, zSig1; - bits64 rem0, rem1, rem2, term0, term1, term2; - floatx80 z; - - aSig = extractFloatx80Frac( a ); - aExp = extractFloatx80Exp( a ); - aSign = extractFloatx80Sign( a ); - bSig = extractFloatx80Frac( b ); - bExp = extractFloatx80Exp( b ); - bSign = extractFloatx80Sign( b ); - zSign = aSign ^ bSign; - if ( aExp == 0x7FFF ) { - if ( (bits64) ( aSig<<1 ) ) return propagateFloatx80NaN( a, b ); - if ( bExp == 0x7FFF ) { - if ( (bits64) ( bSig<<1 ) ) return propagateFloatx80NaN( a, b ); - goto invalid; - } - return packFloatx80( zSign, 0x7FFF, LIT64( 0x8000000000000000 ) ); - } - if ( bExp == 0x7FFF ) { - if ( (bits64) ( bSig<<1 ) ) return propagateFloatx80NaN( a, b ); - return packFloatx80( zSign, 0, 0 ); - } - if ( bExp == 0 ) { - if ( bSig == 0 ) { - if ( ( aExp | aSig ) == 0 ) { - invalid: - float_raise( float_flag_invalid ); - z.low = floatx80_default_nan_low; - z.high = floatx80_default_nan_high; - return z; - } - float_raise( float_flag_divbyzero ); - return packFloatx80( zSign, 0x7FFF, LIT64( 0x8000000000000000 ) ); - } - normalizeFloatx80Subnormal( bSig, &bExp, &bSig ); - } - if ( aExp == 0 ) { - if ( aSig == 0 ) return packFloatx80( zSign, 0, 0 ); - normalizeFloatx80Subnormal( aSig, &aExp, &aSig ); - } - zExp = aExp - bExp + 0x3FFE; - rem1 = 0; - if ( bSig <= aSig ) { - shift128Right( aSig, 0, 1, &aSig, &rem1 ); - ++zExp; - } - zSig0 = estimateDiv128To64( aSig, rem1, bSig ); - mul64To128( bSig, zSig0, &term0, &term1 ); - sub128( aSig, rem1, term0, term1, &rem0, &rem1 ); - while ( (sbits64) rem0 < 0 ) { - --zSig0; - add128( rem0, rem1, 0, bSig, &rem0, &rem1 ); - } - zSig1 = estimateDiv128To64( rem1, 0, bSig ); - if ( (bits64) ( zSig1<<1 ) <= 8 ) { - mul64To128( bSig, zSig1, &term1, &term2 ); - sub128( rem1, 0, term1, term2, &rem1, &rem2 ); - while ( (sbits64) rem1 < 0 ) { - --zSig1; - add128( rem1, rem2, 0, bSig, &rem1, &rem2 ); - } - zSig1 |= ( ( rem1 | rem2 ) != 0 ); - } - return - roundAndPackFloatx80( - floatx80_rounding_precision, zSign, zExp, zSig0, zSig1 ); - -} - -/* -------------------------------------------------------------------------------- -Returns the remainder of the extended double-precision floating-point value -`a' with respect to the corresponding value `b'. The operation is performed -according to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -floatx80 floatx80_rem( floatx80 a, floatx80 b ) -{ - flag aSign, bSign, zSign; - int32 aExp, bExp, expDiff; - bits64 aSig0, aSig1, bSig; - bits64 q, term0, term1, alternateASig0, alternateASig1; - floatx80 z; - - aSig0 = extractFloatx80Frac( a ); - aExp = extractFloatx80Exp( a ); - aSign = extractFloatx80Sign( a ); - bSig = extractFloatx80Frac( b ); - bExp = extractFloatx80Exp( b ); - bSign = extractFloatx80Sign( b ); - if ( aExp == 0x7FFF ) { - if ( (bits64) ( aSig0<<1 ) - || ( ( bExp == 0x7FFF ) && (bits64) ( bSig<<1 ) ) ) { - return propagateFloatx80NaN( a, b ); - } - goto invalid; - } - if ( bExp == 0x7FFF ) { - if ( (bits64) ( bSig<<1 ) ) return propagateFloatx80NaN( a, b ); - return a; - } - if ( bExp == 0 ) { - if ( bSig == 0 ) { - invalid: - float_raise( float_flag_invalid ); - z.low = floatx80_default_nan_low; - z.high = floatx80_default_nan_high; - return z; - } - normalizeFloatx80Subnormal( bSig, &bExp, &bSig ); - } - if ( aExp == 0 ) { - if ( (bits64) ( aSig0<<1 ) == 0 ) return a; - normalizeFloatx80Subnormal( aSig0, &aExp, &aSig0 ); - } - bSig |= LIT64( 0x8000000000000000 ); - zSign = aSign; - expDiff = aExp - bExp; - aSig1 = 0; - if ( expDiff < 0 ) { - if ( expDiff < -1 ) return a; - shift128Right( aSig0, 0, 1, &aSig0, &aSig1 ); - expDiff = 0; - } - q = ( bSig <= aSig0 ); - if ( q ) aSig0 -= bSig; - expDiff -= 64; - while ( 0 < expDiff ) { - q = estimateDiv128To64( aSig0, aSig1, bSig ); - q = ( 2 < q ) ? q - 2 : 0; - mul64To128( bSig, q, &term0, &term1 ); - sub128( aSig0, aSig1, term0, term1, &aSig0, &aSig1 ); - shortShift128Left( aSig0, aSig1, 62, &aSig0, &aSig1 ); - expDiff -= 62; - } - expDiff += 64; - if ( 0 < expDiff ) { - q = estimateDiv128To64( aSig0, aSig1, bSig ); - q = ( 2 < q ) ? q - 2 : 0; - q >>= 64 - expDiff; - mul64To128( bSig, q<<( 64 - expDiff ), &term0, &term1 ); - sub128( aSig0, aSig1, term0, term1, &aSig0, &aSig1 ); - shortShift128Left( 0, bSig, 64 - expDiff, &term0, &term1 ); - while ( le128( term0, term1, aSig0, aSig1 ) ) { - ++q; - sub128( aSig0, aSig1, term0, term1, &aSig0, &aSig1 ); - } - } - else { - term1 = 0; - term0 = bSig; - } - sub128( term0, term1, aSig0, aSig1, &alternateASig0, &alternateASig1 ); - if ( lt128( alternateASig0, alternateASig1, aSig0, aSig1 ) - || ( eq128( alternateASig0, alternateASig1, aSig0, aSig1 ) - && ( q & 1 ) ) - ) { - aSig0 = alternateASig0; - aSig1 = alternateASig1; - zSign = ! zSign; - } - return - normalizeRoundAndPackFloatx80( - 80, zSign, bExp + expDiff, aSig0, aSig1 ); - -} - -/* -------------------------------------------------------------------------------- -Returns the square root of the extended double-precision floating-point -value `a'. The operation is performed according to the IEC/IEEE Standard -for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -floatx80 floatx80_sqrt( floatx80 a ) -{ - flag aSign; - int32 aExp, zExp; - bits64 aSig0, aSig1, zSig0, zSig1; - bits64 rem0, rem1, rem2, rem3, term0, term1, term2, term3; - bits64 shiftedRem0, shiftedRem1; - floatx80 z; - - aSig0 = extractFloatx80Frac( a ); - aExp = extractFloatx80Exp( a ); - aSign = extractFloatx80Sign( a ); - if ( aExp == 0x7FFF ) { - if ( (bits64) ( aSig0<<1 ) ) return propagateFloatx80NaN( a, a ); - if ( ! aSign ) return a; - goto invalid; - } - if ( aSign ) { - if ( ( aExp | aSig0 ) == 0 ) return a; - invalid: - float_raise( float_flag_invalid ); - z.low = floatx80_default_nan_low; - z.high = floatx80_default_nan_high; - return z; - } - if ( aExp == 0 ) { - if ( aSig0 == 0 ) return packFloatx80( 0, 0, 0 ); - normalizeFloatx80Subnormal( aSig0, &aExp, &aSig0 ); - } - zExp = ( ( aExp - 0x3FFF )>>1 ) + 0x3FFF; - zSig0 = estimateSqrt32( aExp, aSig0>>32 ); - zSig0 <<= 31; - aSig1 = 0; - shift128Right( aSig0, 0, ( aExp & 1 ) + 2, &aSig0, &aSig1 ); - zSig0 = estimateDiv128To64( aSig0, aSig1, zSig0 ) + zSig0 + 4; - if ( 0 <= (sbits64) zSig0 ) zSig0 = LIT64( 0xFFFFFFFFFFFFFFFF ); - shortShift128Left( aSig0, aSig1, 2, &aSig0, &aSig1 ); - mul64To128( zSig0, zSig0, &term0, &term1 ); - sub128( aSig0, aSig1, term0, term1, &rem0, &rem1 ); - while ( (sbits64) rem0 < 0 ) { - --zSig0; - shortShift128Left( 0, zSig0, 1, &term0, &term1 ); - term1 |= 1; - add128( rem0, rem1, term0, term1, &rem0, &rem1 ); - } - shortShift128Left( rem0, rem1, 63, &shiftedRem0, &shiftedRem1 ); - zSig1 = estimateDiv128To64( shiftedRem0, shiftedRem1, zSig0 ); - if ( (bits64) ( zSig1<<1 ) <= 10 ) { - if ( zSig1 == 0 ) zSig1 = 1; - mul64To128( zSig0, zSig1, &term1, &term2 ); - shortShift128Left( term1, term2, 1, &term1, &term2 ); - sub128( rem1, 0, term1, term2, &rem1, &rem2 ); - mul64To128( zSig1, zSig1, &term2, &term3 ); - sub192( rem1, rem2, 0, 0, term2, term3, &rem1, &rem2, &rem3 ); - while ( (sbits64) rem1 < 0 ) { - --zSig1; - shortShift192Left( 0, zSig0, zSig1, 1, &term1, &term2, &term3 ); - term3 |= 1; - add192( - rem1, rem2, rem3, term1, term2, term3, &rem1, &rem2, &rem3 ); - } - zSig1 |= ( ( rem1 | rem2 | rem3 ) != 0 ); - } - return - roundAndPackFloatx80( - floatx80_rounding_precision, 0, zExp, zSig0, zSig1 ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the extended double-precision floating-point value `a' is -equal to the corresponding value `b', and 0 otherwise. The comparison is -performed according to the IEC/IEEE Standard for Binary Floating-point -Arithmetic. -------------------------------------------------------------------------------- -*/ -flag floatx80_eq( floatx80 a, floatx80 b ) -{ - - if ( ( ( extractFloatx80Exp( a ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( a )<<1 ) ) - || ( ( extractFloatx80Exp( b ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( b )<<1 ) ) - ) { - if ( floatx80_is_signaling_nan( a ) - || floatx80_is_signaling_nan( b ) ) { - float_raise( float_flag_invalid ); - } - return 0; - } - return - ( a.low == b.low ) - && ( ( a.high == b.high ) - || ( ( a.low == 0 ) - && ( (bits16) ( ( a.high | b.high )<<1 ) == 0 ) ) - ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the extended double-precision floating-point value `a' is -less than or equal to the corresponding value `b', and 0 otherwise. The -comparison is performed according to the IEC/IEEE Standard for Binary -Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag floatx80_le( floatx80 a, floatx80 b ) -{ - flag aSign, bSign; - - if ( ( ( extractFloatx80Exp( a ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( a )<<1 ) ) - || ( ( extractFloatx80Exp( b ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( b )<<1 ) ) - ) { - float_raise( float_flag_invalid ); - return 0; - } - aSign = extractFloatx80Sign( a ); - bSign = extractFloatx80Sign( b ); - if ( aSign != bSign ) { - return - aSign - || ( ( ( (bits16) ( ( a.high | b.high )<<1 ) ) | a.low | b.low ) - == 0 ); - } - return - aSign ? le128( b.high, b.low, a.high, a.low ) - : le128( a.high, a.low, b.high, b.low ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the extended double-precision floating-point value `a' is -less than the corresponding value `b', and 0 otherwise. The comparison -is performed according to the IEC/IEEE Standard for Binary Floating-point -Arithmetic. -------------------------------------------------------------------------------- -*/ -flag floatx80_lt( floatx80 a, floatx80 b ) -{ - flag aSign, bSign; - - if ( ( ( extractFloatx80Exp( a ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( a )<<1 ) ) - || ( ( extractFloatx80Exp( b ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( b )<<1 ) ) - ) { - float_raise( float_flag_invalid ); - return 0; - } - aSign = extractFloatx80Sign( a ); - bSign = extractFloatx80Sign( b ); - if ( aSign != bSign ) { - return - aSign - && ( ( ( (bits16) ( ( a.high | b.high )<<1 ) ) | a.low | b.low ) - != 0 ); - } - return - aSign ? lt128( b.high, b.low, a.high, a.low ) - : lt128( a.high, a.low, b.high, b.low ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the extended double-precision floating-point value `a' is equal -to the corresponding value `b', and 0 otherwise. The invalid exception is -raised if either operand is a NaN. Otherwise, the comparison is performed -according to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag floatx80_eq_signaling( floatx80 a, floatx80 b ) -{ - - if ( ( ( extractFloatx80Exp( a ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( a )<<1 ) ) - || ( ( extractFloatx80Exp( b ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( b )<<1 ) ) - ) { - float_raise( float_flag_invalid ); - return 0; - } - return - ( a.low == b.low ) - && ( ( a.high == b.high ) - || ( ( a.low == 0 ) - && ( (bits16) ( ( a.high | b.high )<<1 ) == 0 ) ) - ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the extended double-precision floating-point value `a' is less -than or equal to the corresponding value `b', and 0 otherwise. Quiet NaNs -do not cause an exception. Otherwise, the comparison is performed according -to the IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag floatx80_le_quiet( floatx80 a, floatx80 b ) -{ - flag aSign, bSign; - - if ( ( ( extractFloatx80Exp( a ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( a )<<1 ) ) - || ( ( extractFloatx80Exp( b ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( b )<<1 ) ) - ) { - if ( floatx80_is_signaling_nan( a ) - || floatx80_is_signaling_nan( b ) ) { - float_raise( float_flag_invalid ); - } - return 0; - } - aSign = extractFloatx80Sign( a ); - bSign = extractFloatx80Sign( b ); - if ( aSign != bSign ) { - return - aSign - || ( ( ( (bits16) ( ( a.high | b.high )<<1 ) ) | a.low | b.low ) - == 0 ); - } - return - aSign ? le128( b.high, b.low, a.high, a.low ) - : le128( a.high, a.low, b.high, b.low ); - -} - -/* -------------------------------------------------------------------------------- -Returns 1 if the extended double-precision floating-point value `a' is less -than the corresponding value `b', and 0 otherwise. Quiet NaNs do not cause -an exception. Otherwise, the comparison is performed according to the -IEC/IEEE Standard for Binary Floating-point Arithmetic. -------------------------------------------------------------------------------- -*/ -flag floatx80_lt_quiet( floatx80 a, floatx80 b ) -{ - flag aSign, bSign; - - if ( ( ( extractFloatx80Exp( a ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( a )<<1 ) ) - || ( ( extractFloatx80Exp( b ) == 0x7FFF ) - && (bits64) ( extractFloatx80Frac( b )<<1 ) ) - ) { - if ( floatx80_is_signaling_nan( a ) - || floatx80_is_signaling_nan( b ) ) { - float_raise( float_flag_invalid ); - } - return 0; - } - aSign = extractFloatx80Sign( a ); - bSign = extractFloatx80Sign( b ); - if ( aSign != bSign ) { - return - aSign - && ( ( ( (bits16) ( ( a.high | b.high )<<1 ) ) | a.low | b.low ) - != 0 ); - } - return - aSign ? lt128( b.high, b.low, a.high, a.low ) - : lt128( a.high, a.low, b.high, b.low ); - -} - -#endif - diff --git a/target-arm/nwfpe/softfloat.h b/target-arm/nwfpe/softfloat.h deleted file mode 100644 index 22c2193..0000000 --- a/target-arm/nwfpe/softfloat.h +++ /dev/null @@ -1,232 +0,0 @@ - -/* -=============================================================================== - -This C header file is part of the SoftFloat IEC/IEEE Floating-point -Arithmetic Package, Release 2. - -Written by John R. Hauser. This work was made possible in part by the -International Computer Science Institute, located at Suite 600, 1947 Center -Street, Berkeley, California 94704. Funding was partially provided by the -National Science Foundation under grant MIP-9311980. The original version -of this code was written as part of a project to build a fixed-point vector -processor in collaboration with the University of California at Berkeley, -overseen by Profs. Nelson Morgan and John Wawrzynek. More information -is available through the Web page `http://HTTP.CS.Berkeley.EDU/~jhauser/ -arithmetic/softfloat.html'. - -THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort -has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT -TIMES RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO -PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY -AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE. - -Derivative works are acceptable, even for commercial purposes, so long as -(1) they include prominent notice that the work is derivative, and (2) they -include prominent notice akin to these three paragraphs for those parts of -this code that are retained. - -=============================================================================== -*/ - -#ifndef __SOFTFLOAT_H__ -#define __SOFTFLOAT_H__ - -/* -------------------------------------------------------------------------------- -The macro `FLOATX80' must be defined to enable the extended double-precision -floating-point format `floatx80'. If this macro is not defined, the -`floatx80' type will not be defined, and none of the functions that either -input or output the `floatx80' type will be defined. -------------------------------------------------------------------------------- -*/ -#define FLOATX80 - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE floating-point types. -------------------------------------------------------------------------------- -*/ -typedef unsigned long int float32; -typedef unsigned long long float64; -typedef struct { - unsigned short high; - unsigned long long low; -} floatx80; - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE floating-point underflow tininess-detection mode. -------------------------------------------------------------------------------- -*/ -extern signed char float_detect_tininess; -enum { - float_tininess_after_rounding = 0, - float_tininess_before_rounding = 1 -}; - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE floating-point rounding mode. -------------------------------------------------------------------------------- -*/ -extern signed char float_rounding_mode; -enum { - float_round_nearest_even = 0, - float_round_to_zero = 1, - float_round_down = 2, - float_round_up = 3 -}; - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE floating-point exception flags. -------------------------------------------------------------------------------- -extern signed char float_exception_flags; -enum { - float_flag_inexact = 1, - float_flag_underflow = 2, - float_flag_overflow = 4, - float_flag_divbyzero = 8, - float_flag_invalid = 16 -}; - -ScottB: November 4, 1998 -Changed the enumeration to match the bit order in the FPA11. -*/ - -extern signed char float_exception_flags; -enum { - float_flag_invalid = 1, - float_flag_divbyzero = 2, - float_flag_overflow = 4, - float_flag_underflow = 8, - float_flag_inexact = 16 -}; - -/* -------------------------------------------------------------------------------- -Routine to raise any or all of the software IEC/IEEE floating-point -exception flags. -------------------------------------------------------------------------------- -*/ -void float_raise( signed char ); - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE integer-to-floating-point conversion routines. -------------------------------------------------------------------------------- -*/ -float32 int32_to_float32( signed int ); -float64 int32_to_float64( signed int ); -#ifdef FLOATX80 -floatx80 int32_to_floatx80( signed int ); -#endif - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE single-precision conversion routines. -------------------------------------------------------------------------------- -*/ -signed int float32_to_int32( float32 ); -signed int float32_to_int32_round_to_zero( float32 ); -float64 float32_to_float64( float32 ); -#ifdef FLOATX80 -floatx80 float32_to_floatx80( float32 ); -#endif - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE single-precision operations. -------------------------------------------------------------------------------- -*/ -float32 float32_round_to_int( float32 ); -float32 float32_add( float32, float32 ); -float32 float32_sub( float32, float32 ); -float32 float32_mul( float32, float32 ); -float32 float32_div( float32, float32 ); -float32 float32_rem( float32, float32 ); -float32 float32_sqrt( float32 ); -char float32_eq( float32, float32 ); -char float32_le( float32, float32 ); -char float32_lt( float32, float32 ); -char float32_eq_signaling( float32, float32 ); -char float32_le_quiet( float32, float32 ); -char float32_lt_quiet( float32, float32 ); -char float32_is_signaling_nan( float32 ); - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE double-precision conversion routines. -------------------------------------------------------------------------------- -*/ -signed int float64_to_int32( float64 ); -signed int float64_to_int32_round_to_zero( float64 ); -float32 float64_to_float32( float64 ); -#ifdef FLOATX80 -floatx80 float64_to_floatx80( float64 ); -#endif - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE double-precision operations. -------------------------------------------------------------------------------- -*/ -float64 float64_round_to_int( float64 ); -float64 float64_add( float64, float64 ); -float64 float64_sub( float64, float64 ); -float64 float64_mul( float64, float64 ); -float64 float64_div( float64, float64 ); -float64 float64_rem( float64, float64 ); -float64 float64_sqrt( float64 ); -char float64_eq( float64, float64 ); -char float64_le( float64, float64 ); -char float64_lt( float64, float64 ); -char float64_eq_signaling( float64, float64 ); -char float64_le_quiet( float64, float64 ); -char float64_lt_quiet( float64, float64 ); -char float64_is_signaling_nan( float64 ); - -#ifdef FLOATX80 - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE extended double-precision conversion routines. -------------------------------------------------------------------------------- -*/ -signed int floatx80_to_int32( floatx80 ); -signed int floatx80_to_int32_round_to_zero( floatx80 ); -float32 floatx80_to_float32( floatx80 ); -float64 floatx80_to_float64( floatx80 ); - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE extended double-precision rounding precision. Valid -values are 32, 64, and 80. -------------------------------------------------------------------------------- -*/ -extern signed char floatx80_rounding_precision; - -/* -------------------------------------------------------------------------------- -Software IEC/IEEE extended double-precision operations. -------------------------------------------------------------------------------- -*/ -floatx80 floatx80_round_to_int( floatx80 ); -floatx80 floatx80_add( floatx80, floatx80 ); -floatx80 floatx80_sub( floatx80, floatx80 ); -floatx80 floatx80_mul( floatx80, floatx80 ); -floatx80 floatx80_div( floatx80, floatx80 ); -floatx80 floatx80_rem( floatx80, floatx80 ); -floatx80 floatx80_sqrt( floatx80 ); -char floatx80_eq( floatx80, floatx80 ); -char floatx80_le( floatx80, floatx80 ); -char floatx80_lt( floatx80, floatx80 ); -char floatx80_eq_signaling( floatx80, floatx80 ); -char floatx80_le_quiet( floatx80, floatx80 ); -char floatx80_lt_quiet( floatx80, floatx80 ); -char floatx80_is_signaling_nan( floatx80 ); - -#endif - -#endif